How a lithium-ion battery wears out

Who it's written for. For someone who doesn't know the industry's terminology and shouldn't have to think in it. Every phenomenon is first described in plain words, and only afterward given its accepted term — so you'll recognize it in reports, standards, and correspondence, while still being able to reason about it in your own words.

How it's organized. The main part moves through locations: anode, cathode, electrolyte, separator, current collectors, casing. Each location has its own breakdown processes. Then come operating modes, a synthesis, and the battery as a system.

⚠ qualifier — the source's wording is imprecise; a corrected version is given alongside

≈ calculation — the figure was obtained by computation, with the full working shown

Question mark — a claim I could not confirm and am holding as unverified

Almost every process below comes down to one question: where do the substance's outermost electrons currently sit, and what follows from that.

Potential doesn't "act on" anything. It says only one thing: whether an electron has somewhere to go. Any transformation of "atom in the lattice ↔ ion in the liquid" consists of two opposing events happening at once, always. For the atom to leave, the electrons it leaves behind must find a place inside the material; for it to return, those electrons must be pulled back out.

Filled to capacity — nowhere to place them, but easy to pull out: the material stays intact. Filled sparsely — easy to place them, hard to pull out: the material dissolves.

A detailed treatment is in section 42, on copper foil.

Before taking the processes apart location by location, you need to see where they all lead. A cell's capacity can drop for exactly three reasons, and they are physically different, even though instruments see the same symptom from all of them.

First: part of the lithium has permanently dropped out of circulation. The lithium is still intact as a substance, but it no longer shuttles between the electrodes — it has bonded with something extraneous and is stuck there for good. This has nothing to do with the electrodes; there's still room in them, there's simply nothing left to carry there. The industry name is loss of lithium inventory, written as LLI in reports.

Second: the graphite has fewer places left. The lithium is intact and mobile, but part of the anode has dropped out of the game: chunks of active material are either destroyed or have lost electrical contact and can no longer be reached. The term is loss of active material at the negative electrode, LAM_NE.

Third: the cathode has fewer places left. The same thing on the other side: part of the cathode material has stopped accepting lithium. The term is LAM_PE, where PE stands for positive electrode.

These three reasons are called degradation modes. Their distinguishing feature is that all three show up on the open-circuit voltage curve: each shifts and deforms it in its own way, and the character of the distortion lets you tell them apart.

The fourth stands apart and doesn't count as a mode: the road has simply gotten harder. The lithium is intact, the sites are intact, but the resistance inside the cell has grown. Under load the voltage sags harder, and the cell hits the lower cutoff threshold before it is actually empty. The instrument records a drop in capacity — but physically the capacity is still there, it just can't be reached at that current.

This distinction matters in practice, not just in theory. A cell with increased resistance will show nearly full capacity at low current and noticeably less at high current. A cell that has lost lithium will show the same reduced capacity regardless of current. Any state-of-health estimate that measures capacity at a single discharge rate cannot tell these two cases apart.

Where this split comes from. The scheme of three modes plus resistance is standard in research practice; the reference work is Birkl and co-authors, Journal of Power Sources, volume 341, 2017, University of Oxford. The distinction there is strict: a degradation process (mechanism) is what happens in a specific part of the cell; a mode is its observable consequence, carrying its own signature in the measured quantities. One process can produce several consequences, and several processes can produce the same one.

A reminder of the starting point. The anode is graphite, operating at a potential of 0.05–0.2 volts — that is, almost at the very bottom of the scale. A charged anode is packed with lithium up to the composition LiC₆: one lithium ion per six carbon atoms. As it fills, the graphene layers spread apart by roughly ten percent, and they come back together as it empties.

Almost all of the anode's troubles follow from these three facts.

A low potential means the anode is the strongest reducing agent in the system. It tends to reduce anything it touches — that is, to hand it an electron. An organic electrolyte cannot exist at such a potential: it is thermodynamically unstable and is bound to decompose.

Breathing by ten percent means mechanical fatigue: a particle that has expanded and contracted a thousand times develops cracks.

Being packed with lithium means that when you try to force in a little more, it has nowhere to go, and it stays on the outside as metal.

The first thing that happens to a new cell on its very first charge: the electrolyte on the graphite surface starts to decompose. It can't be otherwise — at 0.05 volts, organic carbonates are unstable.

But the decomposition doesn't run forever. The breakdown products are solids, and they settle onto the graphite surface as a continuous layer tens of nanometers thick. This layer has a remarkable property: it lets lithium ions through, but not electrons.

This combination is exactly what saves the situation. For the electrolyte to keep decomposing, it needs electrons from the graphite surface. The film blocks them — so the reaction stops on its own. Lithium, meanwhile, keeps moving freely through it in both directions.

The industry name for this layer is the interphase at the boundary between the solid and the electrolyte; in the literature and in reports it is almost always called SEI for short, from solid electrolyte interphase. The abbreviation is used as a proper name in every language, rather than translated — and the same holds in English: SEI is simply what everyone calls it.

What it's made of. The layer is not uniform and splits into two by depth. Closer to the graphite lies the dense inorganic part — lithium fluoride, lithium carbonate, lithium oxide. This is a fine-crystalline, dense substance that insulates well against electrons. On the outside is the loose organic part, made of products of incomplete solvent breakdown — semi-permeable and much less dense.

Why additives go into the electrolyte. If the film were left to form on its own, it would come out loose, uneven, and poorly adherent. So substances are added to the electrolyte that decompose before the main solvent and give the film the right structure — chiefly vinylene carbonate, VC. It reduces first and forms a polymeric component that makes the layer elastic and cohesive. That's also why its share in the electrolyte is only a few percent: it's used up once, during formation.

What formation is. At the factory, a new cell goes through several slow initial charge-discharge cycles under controlled conditions specifically so that this layer comes together correctly. The procedure is called formation; it takes hours or days and accounts for a noticeable share of a cell's manufacturing cost. A poorly formed film condemns the cell to accelerated wear regardless of how it is used afterward.

If the film were a perfect insulator for electrons, the story would end there: the layer would form, the reaction would stop, and the cell would live forever. That's not what happens, and there are several reasons.

Electrons still leak through. Wherever the layer is thin, an electron gets through by quantum tunneling — a particle's ability to end up on the other side of a barrier that, by classical reasoning, it cannot cross. The probability drops sharply with thickness, but it never reaches zero.

The layer is porous. The outer organic part is loose; solvent seeps into its pores and gets closer to the graphite surface, where electrons are within reach.

The layer is partly soluble. Some of its components slowly dissolve back into the electrolyte, exposing bare surface, and the whole process starts over there.

The layer cracks along with the graphite. Every cycle, the particle expands and contracts, while the film on it is rigid. It tears, exposing fresh graphite, and a new film immediately grows on that fresh spot.

The upshot: the film grows for the entire life of the cell. Not fast, but without stopping, and there's no way to halt it by design — only to slow it down.

Every act of film growth takes two things, and both are gone for good.

Lithium. The film consists of lithium compounds — fluoride, carbonate, oxide. The lithium in them is chemically bound and no longer takes part in shuttling. It isn't physically lost; it sits right there, a few nanometers from the graphite, but it has dropped out. This is the main source of the first of the three ways to lose capacity: the lithium inventory shrinks.

An important subtlety. The lithium that went into the film came from the cathode — on the first charge it left the cathode and, in the working sense, never made it to the anode. That's why a cell's capacity after formation is always lower than the calculated value, and this difference — the irreversible first-cycle loss — usually amounts to a few percent, and considerably more for silicon-containing compositions.

Electrolyte. The film is built from solvent and salt. Their amount in the cell is limited and is not replenished. Once enough electrolyte has been consumed, drying-out begins: the pores of the electrodes and separator stop being fully filled, part of the active material loses its ionic connection to the rest of the cell, and this is now a loss of sites, not of lithium.

Resistance. A thickening layer is an extra barrier on the ion's path into the graphite and back out. As it grows, so does the part of the internal resistance tied to the ion crossing that boundary. The cell starts sagging harder under load.

In total, one process produces three consequences at once: lithium loss, rising resistance, and — at later stages, through drying-out — loss of sites. This is typical: the correspondence between processes and consequences is not one to one.

Film growth has a characteristic shape over time, and it follows from the film's own structure.

The thicker the layer, the harder it is for the reacting substances to reach each other through it — an electron going out, solvent going in. So the growth rate is inversely proportional to the thickness already built up. Mathematically, this means the thickness grows as the square root of time.

≈ What this means in practice. If, over the first month of storage, the film took away one unit of capacity, then over four months it will take away two, over nine months three, and over a year about three and a half. Not four, nine, and twelve — their square roots.

This leads to a practical conclusion that often surprises people: a battery loses capacity fastest at the very start of its life, then more and more slowly. The first few percent go in months; the next few go in years. An owner alarmed by a three-percent drop in the first year usually won't see a thirty-percent drop over ten.

This same root law is why wear from simple storage is called calendar aging: it runs on the clock, whether the car is being driven or standing still. It isn't the only component of calendar aging, but it's the main one.

⚠ Where the wording usually breaks down. The root law describes film growth on an undamaged surface. As soon as particle cracking is added, exposing fresh graphite, the dependence stops being a square root and speeds up. So the root law describes a battery sitting idle well, and a hard-working one poorly. The claim "capacity falls as the square root of time" holds only for storage.

The rate of film growth is set by two conditions that the owner controls directly.

Temperature. The reaction is chemical, and its rate obeys the general law of chemistry: it grows not linearly but exponentially with temperature. A rough rule of thumb is that the rate doubles for every ten degrees; the exact figure differs for a given cell, but that's the order of magnitude. This means a battery that sat in the sun all summer ages as much as it would over several years in a temperate climate.

State of charge during storage. The link here is less obvious and builds on what was covered earlier. The anode's potential depends on how full it is with lithium: the fuller the anode, the lower the potential, the stronger the anode acts as a reducing agent, and the faster it decomposes the electrolyte. A fully charged battery means an anode at the very bottom of the potential scale — that is, in its most aggressive state.

Hence the storage rule usually given without explanation: keep it at around half charge. Now you can see where it comes from: at half charge the anode's potential is noticeably higher, and the reaction runs slower. On top of this, the cathode runs its own destructive chemistry at high charge — covered in the part on the cathode.

Combined effect. The two accelerators multiply rather than add. A fully charged battery in the heat ages incomparably faster than a half-charged one in the cool — and the difference is measured in multiples, not percent.

The second major process in the anode is mechanical, and it has nothing to do with film chemistry.

On every charge the graphene layers spread apart to let lithium in, and on discharge they come back together. The whole particle's expansion is about ten percent by volume. For one cycle that's nothing. For a thousand cycles, it's fatigue failure: microcracks form in the particle, they grow, and the particle splits apart.

What matters here is that graphite isn't a solid chunk. The electrode is made from powder: particles a few to a few dozen micrometers in size, each itself made of many crystalline regions glued together. Cracks run along the boundaries of these regions, and the particle gradually crumbles into fragments.

First consequence — new surface. Every fracture exposes clean graphite, uncoated by film and in contact with the electrolyte. A new film immediately grows on it, taking more lithium and more electrolyte. So the mechanical process feeds the chemical one: the more cycles, the more fresh surface, the faster the lithium disappears. This is exactly why the square-root law stops working for a battery that's being worked hard.

Second consequence — loss of contact, covered separately below.

What speeds up cracking. Depth of discharge: a cycle from full to empty gives the full swing of breathing, a cycle within a narrow charge window gives a small one. This is exactly why operating within, say, a twenty-to-eighty-percent range extends battery life by multiples, not by percentage points: the amplitude of mechanical deformation drops, and fatigue failure slows sharply.

Current also matters, but differently: at high current lithium enters the particle unevenly, straining the outside more than the inside, which creates additional internal stresses. Cold amplifies the same effect, because lithium spreads inward more slowly.

A fragment from a split particle may stay in place, or it may shift and lose touch with the electrode's conductive network — the web of carbon black that ties all the particles to the foil.

A fragment that loses contact drops out of the work entirely. Lithium won't come to it: to accept an ion, a particle must simultaneously receive an electron, and now there's nowhere for that electron to come from. The material is physically still there, still intact, but for the cell it no longer exists.

This is the second of the three ways to lose capacity: the graphite has fewer sites left.

Two other processes lead to the same result. The binder polymer holding the mass onto the foil loses elasticity over time and cracks — particles flake off in whole groups. And electrolyte drying-out: a particle with no liquid left around it loses its ionic connection rather than its electronic one, but it drops out of the work just the same.

How this differs from lithium loss in how it shows up. When lithium drops out, the electrodes stay intact, and the cell behaves as healthy, just with lower capacity. When chunks of the anode drop out, the balance between the two electrodes' capacities shifts, and this deforms the voltage curve's shape differently than lithium loss does. Diagnostics — covered in the chapter on measurement — are built on telling these two deformations apart.

The most dangerous process in the anode, and the only one that brings not just wear but fire risk.

Condition for it to occur. During charging, lithium has to get through three things in a row: shed its solvation shell at the boundary, slot in between the graphene layers, and spread into the depth of the particle. Each step takes time. If the charge current delivers lithium to the surface faster than these three steps can absorb it, lithium builds up at the surface.

From there the same equilibrium logic applies that's worked through in detail below using copper as an example (section 42). Here the pair is different — "metallic lithium and its ion" — and its equilibrium point sits at exactly zero on the scale, because the scale is measured from this very pair.

What zero on the lithium scale means. Not a mark on a ruler. Take a piece of metallic lithium in this same electrolyte: some of its atoms are constantly leaving as ions into the liquid, and some ions are constantly settling back as metal. The state where both flows are equal is declared to be zero. In other words, zero is the physical boundary between "it's more favorable for lithium to be an ion in the liquid" and "it's more favorable for lithium to be metal on the surface."

What happens during charging. As long as lithium manages to enter the graphite, the anode's potential stays within its working window of 0.05–0.2 volts — that is, above zero, where metallic lithium is unstable and doesn't form. But the buildup of ions at the surface and resistance along the path pull the local potential down. Once it drops below zero, the balance between the two flows flips: deposition outpaces dissolution, and metallic lithium starts growing on the surface.

Why "more favorable." Turning an ion into metal requires an electron: the Li⁺ ion accepts one and becomes a neutral atom. Whether the reaction proceeds is decided by where the electron sits in the graphite relative to the spot the ion is offering. Below zero, electrons in the graphite end up higher than the offered spot — and start falling into it. Each such drop turns an ion into a metal atom. That's the entire meaning of "more favorable": the electron now has a lower spot to fall into.

The industry name is lithium plating; reports simply call it plating.

What promotes it. Anything that slows entry, and anything that speeds up the supply.

Cold slows entry for two reasons at once: there isn't enough energy to shed the solvation shell, and lithium spreads into the depth of the particle more slowly. This is the main factor, and it explains why fast charging in freezing conditions is so damaging.

High charge current speeds up the supply. A high state of charge means the graphite is already nearly full, with almost nowhere left to accept lithium. A worn cell with a thick film matters too — the film itself creates an extra barrier at the entry point. This last factor makes the process self-accelerating: the older the cell, the milder the conditions under which plating begins.

Numerical thresholds. The specific temperature and current values above and below which plating begins depend on the cell's design — electrode thickness, particle size, electrolyte composition, the balance between anode and cathode capacities. There's no universal threshold; for a given cell it's determined by testing. Any figure given without saying which cell it was obtained for should be treated as a rough guide, not a constant.

Deposited metal has two different fates, and they lead to different consequences.

Part of it returns. If the cell is left to sit after charging, part of the deposited metal manages to dissolve back and enter the graphite — slowly, but on its own. This portion of capacity isn't lost.

Part of it gets cut off. Metallic lithium is extremely reactive and immediately reacts with the electrolyte, growing a film around itself. If this film completely surrounds an island of metal, it ends up electrically isolated from the anode and can no longer return. This is dead lithium — a substance physically present in the cell and completely removed from active service. The industry name is dead lithium.

Part of it grows upward. Metal deposition is uneven: wherever a bump already exists, the field is stronger there, and growth accelerates exactly at that spot. This is how needles form — dendrites. A needle grows from the anode toward the cathode. If it grows far enough to reach the separator and pierce it, the anode and cathode connect directly.

What happens at the puncture. Current flows through the point of contact, limited only by the cell's internal resistance. The contact point heats up. Heating speeds up reactions, they release heat, and the heat speeds up the reactions further. This self-sustaining runaway is exactly thermal runaway. The separator melts shut under overheating, which should stop the process, but once a short circuit has already started, that may not be enough.

A delayed danger. The puncture doesn't have to happen at the moment of charging. A needle can grow gradually, episode by episode, and the short can occur days or weeks after the last cold fast charge — at rest, parked, with no external trigger at all. This makes lithium plating the only process on this list where safety and wear diverge in time: capacity dropped today, but the fire may come later.

There's no direct sensor for this. But a trace is left behind, and it can be found in ordinary voltage measurements.

When a cell is left to rest after charging, the deposited metal starts dissolving and moving back into the graphite. While this is happening, the cell's voltage doesn't behave the way it does during ordinary relaxation: instead of a smooth return to equilibrium, the curve shows a segment with a characteristic kink or plateau. It corresponds to the metal dissolving and differs in shape from a simple concentration equalization.

The practical consequence for this work: seeing this requires voltage records taken at rest after charging, with sufficient sampling rate and long enough duration — tens of minutes. Data captured only while driving contains no trace of it.

Unverified. How reliably this signature can be distinguished on a real vehicle, as opposed to a lab cell, I cannot confirm, and I'm holding it as an open question pending a check against measurements.

A separate process, rare with a healthy electrolyte, but destructive.

A lithium ion in the liquid is surrounded by solvent molecules and must shed them before entering between the layers. If this doesn't happen and the ion enters together with its entourage, what ends up between the graphene layers isn't a small ion but a bulky complex. The layers spread apart not by ten percent but far more, and the graphite exfoliates — literally unravels into separate sheets, irreversibly.

The solvent molecules that end up trapped inside then decompose, releasing gas, which finishes off the destruction and swells the cell.

The industry name is solvent co-intercalation.

What protects against this is precisely the surface film: it's built to let the bare ion through while blocking the complex. This leads to a non-obvious conclusion: solvent composition is chosen not for conductivity but for what kind of film it forms. Ethylene carbonate is present in nearly every electrolyte for exactly this reason — it produces a film that reliably screens out complexes. Propylene carbonate, similar to it in properties and better for cold tolerance, doesn't produce such a film and exfoliates graphite — which is why it isn't used pure with a graphite anode.

Two supporting components of the anode mass that haven't come up yet. Neither stores lithium nor takes part in the reaction, but the electrode doesn't work without them — and both degrade.

The binder. On the anode this is usually a pair of styrene-butadiene rubber and carboxymethyl cellulose, not the fluorine-containing binder used on the cathode.

Why this one specifically. There are four reasons, all practical: rubber-based binder is significantly cheaper; it holds the active material and current collector more firmly; it's processed with water rather than an organic solvent; and it produces a smaller irreversible first-cycle loss, because the cellulose component coats the graphite particles with a thin layer and reduces the exposed surface on which the protective film grows. On top of that, rubber is elastic, and graphite breathes by ten percent on every cycle, so the binder has to keep up with it.

⚠ Correction to the first edition. This section used to say that the fluorine-containing binder on the anode gets reduced and loses fluorine, and that this was the reason for the choice. Checking this did not confirm it: fluorine-containing binder is used on anodes and works there, it's directly compared with the rubber-based one, and at low temperatures it even wins on resistance and lithium diffusion rate. The reason for the choice is the practical considerations listed above, not chemical impossibility.

Separately: the fluorine-containing binder does react with lithiated carbon and release heat at temperatures above 200 °C. But that belongs to thermal runaway (Part VIII), not to normal operation.

How the binder degrades. Three ways. It gradually loses elasticity and stops keeping up with the graphite's breathing — then particles tear away from the mass in groups. It's partly built into the protective film as it grows, meaning it's consumed together with the electrolyte. And it's corroded by hydrogen fluoride arriving from the cathode side (section 29).

Conductive carbon black. Needed so an electron from deep in the electrode can reach the foil: graphite particles sit apart from each other in the mass, and without a conductive network there's no connection between them.

How it does harm. Carbon black has an enormous specific surface area, and all of that surface is in contact with the electrolyte at the anode's potential. So a protective film grows on it too — with all the accompanying consumption of lithium and solvent. More carbon black means better current delivery, and also more irrecoverable lithium loss during formation.

This is one of the reasons the irreversible first-cycle loss differs between cells with the same chemistry: different amounts of carbon black and different specific surface area of the graphite.

A shared consequence. Failure of either one produces the same result — loss of electrical contact, that is, a loss of sites in the anode. It cannot be told apart from contact loss due to particle cracking (section 8) by external measurements.

Common to all seven processes. Four are chemical and run on the clock: film growth, metal deposition, foil dissolution, graphite exfoliation. Three are mechanical and run on cycles: cracking, contact loss, mass detachment. The first group is sped up by heat and state of charge, the second by depth of discharge and current. So a battery sitting idle and a battery being worked hard don't just wear at different speeds — they wear differently.

What this means for operation. Three levers, and all three follow from what's been covered, not from general considerations. Don't keep the battery fully charged without need — this slows film growth. Don't keep it in the heat — the same thing, only stronger. Don't fast-charge it in the cold — this is the only measure against metal deposition, and it's a matter of safety, not just service life.

What's not on this list. None of these processes stops completely, and none reverses. Everything listed can only be slowed down. Claims that some operating procedure can restore the capacity of a worn battery have nothing to do with these processes.

Three conditions, and almost everything else follows from them.

High potential, 3.6–4.2 volts. The cathode is the strongest oxidizing agent in the system. By the logic from section 42.2: its electron states are emptied out, and it hungrily pulls electrons from anything it can reach — including the electrolyte.

Oxygen in the lattice. Graphite has nothing to give up from its own structure — it's made of carbon. The cathode is built from oxygen, and that's its weak point.

A rigid slab structure. The MO₂ slabs are held together by strong metal–oxygen bonds and barely expand as lithium leaves. There's no ten-percent breathing here the way there is with graphite — but mechanical damage still happens, just in a different form (section 22).

For the anode, the central story was the protective film. The cathode has a direct counterpart to it, and it's built considerably worse.

At 4.2 volts the electrolyte is unstable in the opposite direction. At the anode it accepted electrons and got reduced; at the cathode it loses them — it gets oxidized. The oxidation products settle on the particle surfaces as a film. Named by analogy: the interphase at the cathode, written in reports as CEI, cathode electrolyte interphase.

How it differs from the anode's film — and this is the key point. The film on graphite is self-limiting: it doesn't conduct electrons, so decomposition beneath it stops. The cathode film does this job poorly — it's thinner, less dense, partly soluble, and doesn't form a continuous barrier. Electrolyte oxidation continues for the entire life of the cell.

What this releases. Oxidizing the solvent molecules frees up protons — hydrogen ions. From there they corrode everything in reach: the oxide film on the aluminum foil, the surface of the cathode particles themselves, and they take part in forming hydrogen fluoride. Almost every trouble attributed to hydrogen fluoride below starts here.

A conclusion that matters for understanding the asymmetry between the electrodes: at the anode, the film takes lithium but protects. At the cathode, it takes lithium and doesn't protect — it is itself a source of aggression.

Nickel gives up electrons during charging, moving in sequence through the states Ni²⁺ → Ni³⁺ → Ni⁴⁺. It's the source of capacity — and, as a consequence, the source of four different kinds of trouble.

Ni⁴⁺ is a fierce oxidizer. It doesn't wait for the electrolyte to approach: it pulls electrons straight from solvent molecules right at the particle surface. The more nickel in the composition and the higher the charge, the faster the electrolyte gets oxidized.

Ni²⁺ migrates into lithium sites. Its ionic radius is 0.69 ångström against lithium's 0.76 — close enough to take its place. The site ends up blocked, the channel narrowed.

Nickel dissolves and migrates to the anode, where it damages the protective film (section 21).

Nickel sets the amplitude of the lattice change. The more of it there is, the more the crystal's dimensions change during cycling — and the more the particles crack.

This gives a symmetry that can't be engineered around: high-nickel compositions deliver more capacity and hold up worse over time. This isn't a technology shortfall — it's a direct consequence of the fact that capacity is exactly what nickel's work produces.

Manganese sits in the Mn⁴⁺ state and doesn't give up electrons within the working potential window. Its job is to hold the framework together. And it really does hold it together.

⚠ A mechanism that doesn't belong here. Manganese dissolution, famous from LiMn₂O₄ spinel, proceeds through disproportionation: two trivalent manganese ions turn into a soluble divalent one and a solid tetravalent one. This requires Mn³⁺. In NMC, manganese is predominantly tetravalent, a state that is stable and electrochemically inactive, while Mn³⁺ is present only as a minor impurity — so NMC aging cannot be driven mainly by manganese. At very high cutoff voltages, above 4.4 volts, manganese does start leaving NMC too, but by a different route — through structural breakdown, not through its own instability.

A practical conclusion that goes beyond manganese: degradation mechanisms cannot be carried across chemistries. What's true for LFP or LMO may not hold for NMC, and vice versa. The claim "manganese dissolves," applied to NMC, is a sign of exactly this kind of transfer and needs checking.

At high charge, Co³⁺ turns into Co⁴⁺ — also a strong oxidizer, also driving electrolyte oxidation. And cobalt dissolves.

But at the same time, cobalt is exactly what suppresses nickel's migration into lithium sites, keeping the structure ordered. Removing it to cut cost means getting more cation mixing.

Hence a second fork in composition, besides "nickel versus service life": cobalt is expensive and is itself aggressive at the top of the range, but without it the structure degrades faster.

Oxygen sits in the lattice as the O²⁻ ion — it holds two extra electrons handed to it by the metals. It isn't neutral, and it's precisely its negative charge that attracts the positive metal ions, holding the slab together.

From here it's the same arithmetic worked through for copper in section 42.2, just mirrored.

Leaving. The O²⁻ ion gives its two extra electrons back to the lattice, becomes a neutral atom, and leaves the structure. Two such atoms combine into an O₂ molecule — a gas.

Staying. The reverse: an atom takes two electrons from the lattice, becomes an ion, and stays put.

The question is the same: is there anywhere to put two electrons. At low charge, the cathode's states are full, there's nowhere to put them — oxygen stays. At high charge, the states are emptied out — oxygen leaves.

Exactly when it starts — and this is more precise than it sounds. Oxygen release begins once about 80–81 percent of the lithium has been extracted. And — this matters for understanding the mechanism — the process is governed not by potential but specifically by the fraction of lithium removed: it starts as soon as that fraction is reached, whatever voltage that happens to correspond to. The shift in the observed onset voltage with temperature is explained purely by differences in overpotential, not by a shift in the threshold itself.

Numbers by composition — and these matter more than the general rule. The onset of oxygen release for NMC811 is around 4.3 volts versus metallic lithium, which in a full cell against graphite corresponds to roughly 4.2 volts at the terminals. For NMC111 and NMC622 it's around 4.7 volts versus lithium — that is, roughly 4.6 volts at the terminals.

This gives the precise reason high-nickel compositions are more dangerous: for NMC111 and 622, at the standard upper limit of 4.2 volts, no oxygen is released at all — the margin is about four tenths of a volt; for NMC811, the onset of release coincides with the upper working limit. The same charging regime means something different for the two compositions: for one it's operation with margin to spare, for the other it's operation right on the edge.

⚠ Correction to the first edition. This section used to say: "once more than half the lithium has been pumped out, nickel is exhausted, and further capacity is drawn from oxygen," with the limit stated as "don't extract more than 50–60 percent of the lithium." The figure of 50–60 came from the source material and wasn't confirmed: the threshold for oxygen release is around 80 percent extraction. A limit on lithium extraction for NMC does genuinely exist and is tied to structural instability at high extraction levels — NMC's theoretical specific capacity is around 275 mAh/g, and the practical figure is considerably lower for exactly this reason — but that's a different limit, one that doesn't coincide with the oxygen-release threshold.

Three consequences, all bad. The gas swells the cell: pressure builds in a cylindrical one, a pouch cell visibly bulges. Oxygen plus organic electrolyte is a ready-made combustible mixture that needs no outside air; that's why a burning lithium-ion battery can't be put out by cutting off the air supply — it carries its own oxidizer. And a slab that has lost some of its supports rearranges itself (section 23).

This is the only place in the whole cell where wear and catastrophe are the same process at different stages. At the anode they're separate: film growth eats capacity, metal deposition creates short-circuit risk — different mechanisms. In the cathode, oxygen release simultaneously reduces capacity and sets the stage for a fire.

Unverified. The specific potential at which oxygen release begins for a given NMC composition is something I have not confirmed against primary sources. The mechanism and its threshold nature are established; the numerical boundary for a specific chemistry needs checking.

Specific to high-nickel compositions. During manufacturing, lithium is added in excess, and part of it remains on the particle surface as lithium carbonate and lithium hydroxide.

These compounds are hygroscopic — they pull moisture from the air while the electrodes are stored before assembly. Moisture plus the LiPF₆ salt produces hydrogen fluoride. The same compounds also react with the electrolyte, releasing carbon dioxide, which is why the cell can swell within the very first cycles.

A consequence that isn't obvious from the operating side: part of a high-nickel cell's service life is determined not by how it's used, but by what atmosphere the electrodes were stored in at the factory.

I said the slabs are rigid and barely expand. That's true for a single slab, but not for the particle as a whole.

A cathode particle isn't a monolith. It's sintered from many small crystalline grains oriented in different directions. As lithium leaves, each grain changes size slightly, in its own direction. Neighboring grains pull on each other, and stresses build up at the boundaries. Over hundreds of cycles the boundaries pull apart — the particle cracks from the inside, without its outer size visibly changing.

First consequence — new surface. A crack exposes fresh material in contact with the electrolyte. Everything described above starts on it immediately: electrolyte oxidation, rearrangement into rock salt, metal dissolution. Mechanics feeds chemistry — exactly as at the anode.

Second — loss of contact. A grain that has broken off loses its connection to the conductive network and drops out of the work.

What speeds it up. The swing of the cycle: the more lithium leaves and re-enters, the more the grains' sizes diverge. And nickel content: the more of it there is, the more the lattice parameters change.

The role of current. This is also where fast charging's harm to the cathode comes from. At high current, lithium leaves the particle unevenly — first from the surface, then from the depth — creating a stress gradient inside the particle on top of the intergranular one. This gives an important distinction: fast charging harms the two electrodes differently. For the anode, it's the risk of metallic lithium deposition. For the cathode, it's mechanical fracture of the particles.

The industry's engineering answer is single-crystal particles, which have no grain boundaries and nothing to pull apart. They last longer and deliver current less well.

Once oxygen has left, the freed-up metal ions — nickel above all — shift into the sites where lithium used to be.

The structure passes through three states:

Layered — the original, working state. Lithium moves freely in two dimensions.

Spinel — intermediate. Some of the metals have moved into the lithium layers, the channels have narrowed, and lithium moves less freely.

Rock salt — the end state. Metals and oxygen are arranged interspersed, and there are no lithium layers left at all. Lithium essentially cannot pass through a structure like this.

The transformation is irreversible: the metals don't migrate back.

Where it happens. It starts at the particle surface — that's where the electrolyte contact is, and where everything most aggressive takes place. The rock-salt layer builds up from the outside and advances inward. Its thickness is substantially greater than one might expect: measured at up to 100 nanometers after 500 cycles at 80 °C. At lower temperatures growth is slower, but the order of magnitude is tens of nanometers, not just a few.

Why a thin layer causes so much damage. It doesn't take the material away — it locks it up. The particle's core is intact, the lithium is there, sites are free, but lithium can no longer pass in and out: there's a crust on the outside it cannot get through. The material is present but behaves as if it were absent.

This is a loss of sites in the cathode combined with a sharp rise in resistance at the same time, because every ion now has to fight its way through the crust.

A mechanism that links the cathode directly to the anode.

Why they leave. Two reasons that add together. The first is the same filling arithmetic: at high potential, an ion leaving the lattice becomes accessible. The second is chemical: hydrogen fluoride corrodes the oxide directly.

What happens next. The dissolved metal ion drifts through the separator toward the anode. There the potential is low — and by the logic already covered, this means the ion readily accepts electrons and settles as metal. It deposits right inside the protective film on the graphite.

Why this is poison. The film worked because it didn't conduct electrons. A metallic inclusion inside it does. At that spot the electrolyte gets access to electrons again, decomposes, the film grows — and the growth is no longer self-limiting.

The upshot: damage to the cathode speeds up wear at the anode. The cell loses lithium faster than either electrode's condition alone would predict. This is exactly the kind of case for which the chapters aren't kept strictly separated by location.

The binder holds the active mass onto the foil. Hydrogen fluoride corrodes it, it oxidizes at high potential, and it loses elasticity over time. The result is particles flaking off in groups and loss of contact. The same process as at the anode, but in a more aggressive environment.

Conductive carbon black is needed so an electron from deep in the electrode can reach the foil. But it has an enormous specific surface area, and it itself serves as a platform for electrolyte oxidation. More carbon black means better current delivery and faster electrolyte decomposition.

For the cathode, it isn't only the top that's harmful. On deep discharge, NMC ends up holding more lithium than its original composition called for: the lattice becomes overloaded, and this too irreversibly damages the surface.

This is why the working window is bounded on both sides, not just from above, and the lower cutoff of around three volts protects not only the copper foil at the anode (section 42) but the cathode itself as well.

28.1. What "the cathode wears out from state of charge" means

For the owner, charge is a number on the dashboard. For the cathode, it's how much lithium has left it — and so how empty its electron states are, that is, its potential. "High charge harms the cathode" and "the cathode at high charge pulls electrons from everything around it more strongly" are the same statement.

The relationship is exponential, not proportional. The rate of the reaction in which an electron crosses the boundary grows by an order of magnitude for roughly every one hundred to one hundred fifty millivolts — the same relationship as for copper in section 42.3, and that's not a coincidence, it's the same mechanism. The difference between 4.0 and 4.2 volts is two hundred millivolts, or one and a half to two decades: electrolyte oxidation at the cathode surface runs roughly fifty times faster. Not twenty percent faster, not twice as fast. This is exactly why the upper limit is set as a hard boundary, not a recommendation.

(⚠ The first edition said "sixty millivolts per decade" here and, as a result, "a thousand times." Checking showed that the baseline theoretical slope for single-electron transfer is around 120 millivolts per decade, with observed values of 30–240. The estimate has been corrected from a thousandfold to roughly fiftyfold. The qualitative conclusion still holds; the scale had been overstated by about twenty times.)

Why the last few percent of charge cost the most. Two things add together. The voltage curve is steeper at the top: going from 50 to 60 percent adds roughly fifty millivolts to the cathode's potential, while going from 90 to 100 adds about twice that. And this larger increase gets raised to a power. The upshot: the last ten percent of charge doesn't cost ten percent of the service life — it costs several times more. This is the physical basis for the advice to charge only to eighty percent.

Element by element. For the cathode film, the oxidation rate rises, and on top of that, at high potential, solvent components that don't oxidize at all lower in the range come into play. For nickel and cobalt the link isn't just about speed: Ni⁴⁺ and Co⁴⁺ — the most aggressive states — exist only at the top of the range; at 3.7 volts there's practically none of them. For oxygen the link is threshold-based: as long as nickel is still able to give up electrons, oxygen is left alone. The rearrangement into rock salt inherits oxygen's threshold behavior. Metal dissolution speeds up in two ways at once — directly through potential, and through a larger yield of hydrogen fluoride. The aluminum foil passes through its point of best protection around four volts and operates beyond that limit from there on.

What doesn't depend on state of charge. Microcracks depend not on the level but on the swing: a cycle from 80 to 90 percent produces a small divergence in grain sizes, a cycle from zero to a hundred produces the full amount. A battery sitting at one hundred percent and not working won't crack at all, even though every one of its other processes is running at maximum. Residual lithium is a manufacturing question. In total, eight of the table's ten rows are governed by state of charge, and two are not.

Three different things that blur together into the word "charge." The maximum reached matters for threshold processes (oxygen release, film breakdown on aluminum): what counts is whether the line was crossed or not. Time spent at high charge matters for rate-based processes: charging to a hundred and driving off right away isn't the same as charging to a hundred and leaving it overnight, even though the maximum is identical. Cycle swing matters only for the mechanics. A practical distinction: charging to a hundred right before a long trip with an immediate departure is acceptable; sitting at a hundred percent parked is pure damage with no benefit.

28.2. What heat does

The general law. The rate of a chemical reaction grows exponentially with temperature; the standard approximation is doubling for every ten degrees. This applies to every chemical process in the cathode without exception.

On top of that — four things specific to the cathode. Heat speeds up hydrolysis of the LiPF₆ salt by trace water, and that is a source of hydrogen fluoride: more hydrogen fluoride means faster metal dissolution, faster corrosion of the binder, and worse protection for the aluminum. Heat worsens the fluoride film's ability to re-heal on the foil after local damage. Heat speeds up diffusion in the solid, and nickel's migration into lithium sites is exactly that — diffusion. And heat generates itself: a worn cell with increased resistance heats up more under load, the heating speeds up wear — a feedback loop.

Why charge and heat multiply rather than add. These are two independent factors. A rough estimate: going from 4.0 to 4.2 volts is about a fiftyfold increase, heating from 25 to 45 degrees is about a fourfold increase, together roughly two hundredfold. A battery sitting fully charged in the summer sun doesn't wear "a bit faster" than a half-charged one in a cool garage: the difference is in the hundreds of times, not in percentage points.

28.3. Asymmetry between the electrodes

The anode wears out mainly from time and heat: film growth runs on the clock, regardless of whether the battery is working or sitting idle.

The cathode wears out from state of charge and heat. At fifty percent charge it's nearly calm: oxidation runs three orders of magnitude slower, there's no Ni⁴⁺, no oxygen release, and the aluminum sits in its zone of best protection.

The recommendation to store at around half charge works for both electrodes, but for different reasons: for the anode it lowers aggressiveness toward the electrolyte, while for the cathode it moves it out of the range where its destructive processes even happen.

Heat is the one factor that harms both sides equally, since it speeds up chemistry everywhere regardless of the mechanism.

Three features, and everything that follows comes from them.

It doesn't wear out on its own — it gets consumed. Graphite and NMC have their own mechanisms of damage: cracks, lattice rearrangement. The electrolyte stays intact until it's used up. It doesn't age from fatigue — it dwindles.

Its amount is not replenished. The cell is sealed; exactly as much was poured in as was poured in at the factory. Every molecule that ends up in the anode's film or gets oxidized at the cathode is lost for good.

It touches everything at once. The electrodes are separated by the separator and never touch each other. The electrolyte touches both, plus both foils, plus the separator. That's why it's the only carrier of damage between components: whatever dissolves at the cathode reaches the anode only through it.

Hydrogen fluoride was mentioned six times in earlier parts without explanation. Here's where it comes from.

The salt LiPF₆ is unstable in the presence of water. Trace amounts of moisture are enough to trigger hydrolysis, and the products are hydrogen fluoride and phosphorus-containing compounds.

Where the water in the cell comes from, if the electrolyte is anhydrous. Four sources, and none of them can be closed off completely.

Residual moisture in the materials: the electrodes and separator are dried before assembly, but perfect drying doesn't exist. Hygroscopic residual lithium on the surface of high-nickel particles (section 21), pulling moisture in from the factory floor's atmosphere. Water as a byproduct of the electrolyte's own decomposition — some oxidation reactions release it. And loss of seal integrity in the casing from damage or aging gaskets.

Why this is a self-sustaining process. Hydrolysis doesn't just run — it's accelerated by what it produces: the decomposition products include compounds that themselves promote further breakdown of the salt. Plus temperature: hydrolysis speeds up exponentially with heat. A cell that has spent time in the heat accumulates hydrogen fluoride that stays with it even after it cools back down.

What hydrogen fluoride does. Gathered from earlier parts: it corrodes the oxide on cathode particles and leaches out transition metals; it corrodes the binder, causing the active mass to lose adhesion to the foil; it dissolves the inorganic part of the anode's protective film, forcing it to regrow at the cost of more lithium; and — the one useful effect — it converts the oxide on the aluminum foil into a fluoride, which is exactly what serves as its working protection (section 43).

Solvent leaves through three channels, all covered earlier; here they're brought together.

Into the anode film. Every thickening of the protective layer is built from solvent decomposition products. It runs for the entire life of the cell, speeds up with heat and high state of charge, and jumps sharply when particles crack.

Into oxidation at the cathode. A constant leak, since the cathode film isn't self-limiting. Speeds up exponentially with potential.

Into gas. Some of the decomposition products are gaseous: carbon dioxide, carbon monoxide, hydrocarbons. They not only swell the cell but also mean the substance has left the liquid phase for good.

This gives a conclusion that isn't obvious from the operating side: cell swelling isn't a standalone failure — it's an indicator of electrolyte consumption. If a cell has swollen, the amount of liquid inside it has dropped by a corresponding amount.

The mechanism that makes the electrolyte worth treating as its own subject.

Electrolyte fills the pores: in the anode mass, in the cathode mass, in the separator. As long as the pores are full, every particle of active material has an ionic connection to the rest of the cell. Once liquid runs short, some of the pores end up dry.

What happens to a dry particle. It's physically intact. There's lithium inside it, sites are free, electrical contact with the conductive network is unbroken. But the lithium ion never reaches it — only the liquid can deliver it. The particle drops out of the work completely.

Why this is a qualitative break, not a quantitative one. Before drying-out, every process took away capacity gradually, a little at a time. Drying-out takes it away as a step: while there's enough liquid, almost everything works; once there isn't enough, whole regions of the electrode drop out at once.

This gives the characteristic shape of the wear curve observed in lithium-ion cells: a slow square-root decline at first, then, after some number of cycles, a sharp bend and a collapse. This bend is, to a large extent, exactly the moment when electrolyte consumption reaches a shortfall. In the literature it's called the knee of the aging curve.

Why this is more dangerous than a simple loss of capacity. Drying-out proceeds unevenly: the regions that got less liquid to begin with dry out first. Current that used to spread over the whole area now flows through what remains wetted. The local current density rises — and high local current density at the anode is exactly the condition for metallic lithium deposition (section 9). In other words, drying-out pushes the cell into a regime where ordinary charging becomes, for individual patches, fast charging with all its consequences.

A distinctive feature among the three ways to lose capacity. Drying-out is the only process that produces a loss of sites on both electrodes at once, because pores dry out on both sides.

A separate line item, because these additives' working life is shorter than the cell's.

Additives are the few percent of the composition that give the anode's protective film its correct structure. They're consumed mainly during formation, but not entirely: the remainder keeps working every time the film gets renewed after particles crack.

Once the additives run out, the film on fresh fractures forms without them — loose, less dense, a worse insulator. Electrolyte decomposition beneath it runs faster.

This is one more contribution to the bend in the wear curve: before the additives run out, every bit of film damage heals well; afterward, it heals poorly.

The electrolyte doesn't just get consumed — it also degrades from what accumulates in it.

Transition-metal ions from the cathode poison the anode's film (section 24). Copper ions, from deep over-discharge (section 42), deposit on the anode, cathode, and separator. Products of its own decomposition — some stay dissolved and speed up further breakdown. Water, released in some reactions, closes the loop with salt hydrolysis.

The overall pattern: as a cell ages, the electrolyte doesn't just shrink in quantity — it also becomes more aggressive in composition. Both factors push in the same direction.

What cold does to the electrolyte. The solvent thickens, conductivity drops, ions move more slowly. This adds to internal resistance and slows the delivery of lithium to the anode surface. At sufficiently low temperatures, solvent components can precipitate out or freeze; carbonate blends are chosen so this doesn't happen within the working range, but the range is finite.

Cold as preservation. By the same law as everything else — the rate doubling for every ten degrees — the difference between minus twenty and plus twenty-five gives roughly a twentyfold slowdown in chemistry. A year of storage in the cold ages a cell about as much as two to three weeks in the warmth. That's a great deal, but it isn't zero: saying the processes are "stopped" is inaccurate — they're slowed by tens of times.

How cold does harm with no current involved at all. Through two things: possible precipitation of solvent components, and uneven thermal contraction of the materials — the metal foil, the polymer separator, and the active mass all contract by different amounts, and stresses build up at the boundaries. Insignificant for one episode, but it accumulates over repeated trips through freezing temperatures.

Charging and discharging in the cold are asymmetric, and this is the key point. During charging, lithium moves into the graphite, and all three steps of its entry are slowed; if the current delivers lithium faster than the graphite can accept it, metal deposition begins. During discharging, lithium moves into the cathode, where metallic lithium is thermodynamically unstable and has nowhere to deposit: discharging in the cold produces a voltage sag and an early cutoff — that is, a drop in power with no damage.

What's harmful isn't the lithium transfer itself, but its direction. You can drive in the cold; you shouldn't charge in it.

How this differs from heat's harm. Heat speeds up processes that happen anyway: the film always grows, the electrolyte always oxidizes. Cold plus charging current triggers a mechanism that doesn't exist at all in the warmth: at twenty-five degrees and moderate current, metallic lithium doesn't deposit — not slowly, but not at all. So "cold speeds up degradation" is an imprecise way to put it. More accurately: cold opens up an additional mechanism of damage while simultaneously slowing down all the others.

A hidden source of charging current. The charging station isn't the only thing that delivers charge. Regenerative braking is charging that appears the instant the vehicle first slows down, with no action from the driver. That's why manufacturers limit regeneration on a cold battery; the driver perceives this as lost braking effectiveness, when it's actually protection against metal deposition.

What matters most in this part. The electrolyte is the only component whose failure is threshold-based. Every other process takes capacity away gradually; a shortage of liquid takes it away as a step, and this explains the bend in the wear curve after which the battery degrades sharply.

What this means for state-of-health estimation. A cell before the bend and a cell after it behave qualitatively differently, and extrapolating across the bend doesn't work. A remaining-life forecast built on data from the first half of life systematically overstates service life — not because of a modeling error, but because the model is describing a different regime.

The separator is the only component of the cell with no active chemistry of its own. It gives up nothing, accepts nothing, takes no part in reactions: a polyethylene or polypropylene film with pores twenty to fifty nanometers across, mechanically keeping the electrodes apart while letting ions through the electrolyte-filled pores.

That's why it doesn't have mechanisms of breakdown — it has damage: what other things do to it.

But one feature makes this part disproportionately important: the separator is the only component whose failure means not a loss of capacity but an accident. Everything covered so far takes away service life gradually. Puncturing the separator connects the electrodes instantly.

The main wear, and it runs constantly. The pores clog with electrolyte decomposition products, fragments of active mass, and deposited metals — everything that forms in the cell and ends up in the liquid eventually settles in the pores.

What this produces. The ionic path between the electrodes narrows — internal resistance rises, specifically the part responsible for transport through the bulk. The cell sags harder under load.

Why this is more dangerous than a simple resistance increase. Clogging happens unevenly: some pores are blocked, others are free, and the current takes the free ones. Local current density in them rises — and that is a direct condition for metallic lithium deposition at the anode. The same result as with drying-out (section 31), and for the same reason: a narrowing of the path, not a change in chemistry.

Puncture by a dendrite. Needles of metallic lithium from the anode, or copper deposits from over-discharge, grow toward the separator. The film's thickness is tens of micrometers, and a needle can get through it (sections 10 and 42).

Puncture by particles. Sharp-edged fragments of active mass, under the stack's constant compression, press through the polymer. The compression comes from the anode expanding during charging — that same ten-percent breathing.

Fatigue from cyclic compression. The electrode stack breathes along with the anode. The polymer compresses and relaxes thousands of times, loses elasticity, and thins out where the pressure is greatest.

Thermal shrinkage. On heating, the film pulls in, and its edges can retreat from the electrode's edge — leaving the anode and cathode facing each other with nothing between them. That's why the separator is always made wider than the electrodes: a margin at the edge in case of shrinkage.

The one function that works as protection rather than as a vulnerability. On heating to around one hundred thirty degrees, the polyethylene softens, the pores seal shut, ionic transport stops — the cell stops working, the reaction halts.

Three caveats, without which the picture is wrong.

Shutdown is irreversible: a cell that has triggered this protection never works again.

It stops ionic current, but not a short circuit. If a conductive bridge has already formed, electrons flow through it, and shutdown doesn't stop them.

On further heating, at roughly one hundred sixty degrees, the softened film loses its shape and tears — the electrodes touch directly. This is why three-layer separators exist: outer layers of polypropylene, which melts later, hold their shape while the inner polyethylene layer shuts the pores. There's a temperature gap between shutdown and tearing, and the entire protection consists in that gap being wide enough.

What matters most here. The separator doesn't break down on its own — it absorbs the consequences of everything else. All of its damage comes from outside: fragments from the electrodes, metal from the electrolyte, heat from the reactions.

A consequence for state-of-health estimation. Of the table's six rows, only one is visible to instruments: pore clogging shows up as rising resistance. The other five — mechanical — give no warning sign before the moment of failure. This is exactly why usage history has to be tracked separately from current measurements: a cell that has been through episodes of cold charging or deep over-discharge carries a risk that current readings won't show.

The active mass is a powder held together by binder. You can't connect a wire to it directly: you need a continuous metal backing that collects electrons from the entire electrode area and carries them out. That's the current collector — a foil ten to twenty micrometers thick, coated with the paste.

It plays no part in working with lithium: it doesn't store it, doesn't accept it, doesn't enter into reactions. Functionally it's an outsider — which makes it all the more interesting that both current collectors can fail.

Why copper on the anode and aluminum on the cathode. Not because of price and not because of conductivity — the two metals are comparable on both counts, and copper is even more expensive. What decides it is the potential the metal has to work at: 0.05–0.2 volts at the anode and 3.6–4.2 at the cathode. No metal is stable across both ranges at once.

Two different ways of surviving. This is the key difference, and it determines how each one fails.

Copper is protected by a state, aluminum by a substance. This is also why prevention differs: copper is safeguarded by a lower voltage cutoff, aluminum by an upper one and by the presence of a fluorine-containing salt.

What they share. Both foils are thin and mechanically vulnerable: tearing during winding, nicks at the edge, fatigue from vibration in automotive use. Both lose cross-section to corrosion, which raises ohmic resistance — the component that sags instantly under current. And both carry the welded joints to the current tabs, covered in the next part.

Copper sits beneath the anode mass and works at its potential. This section is worked through in more detail than the others: it's the easiest place to show how potential governs a substance's fate at all. The same logic is applied later to oxygen in the cathode (section 20) and to aluminum in the next section.

42.1. What physically happens on the copper surface

The copper foil is wetted by the electrolyte. This isn't obvious and is worth spelling out: the graphite mass is applied to the foil as a paste and stays porous after drying, and when the cell is assembled, it's filled with electrolyte that soaks all the way through the pores — right down to the foil. Otherwise the cell wouldn't work: a lithium ion has to reach every graphite particle, and only the liquid delivers it. So the foil isn't isolated from the electrolyte — it's sitting in it.

Two opposing events happen constantly on the wetted surface.

Leaving. A copper atom detaches from the lattice and moves into the liquid. It doesn't take its two outer electrons with it — they stay in the metal. In the liquid it becomes a Cu²⁺ ion, and solvent molecules immediately surround it, stabilizing the charge — the same way they surround a lithium ion. This is exactly why an ion, not a neutral atom, is what leaves for the liquid: a neutral copper atom has nothing to hold onto in an organic solvent.

Returning. A Cu²⁺ ion approaches the surface, sheds its solvent molecules, takes two electrons from the metal, and slots back into the lattice.

Both events happen always and at once, at any potential, for the entire life of the cell. Nothing "switches on." Only the ratio of the rates changes.

42.2. What decides the balance

Look at what each event requires.

Leaving needs a place in the metal for the two spare electrons. Returning needs to pull two electrons out of the metal.

Both depend on the very same thing — how full the metal's electron states are — and depend on it in opposite directions.

When the states are filled to capacity and the topmost electrons sit high: there's nowhere to place two more, so leaving is almost impossible; but pulling out the topmost ones is easy, since they're held weakly, so returning happens readily. Net result: the foil stays intact.

When the states are filled sparsely and the remaining electrons sit deep: placing two more is easy, since there are free spots; but pulling them out is hard, since deep electrons are held tightly. Net result: the foil dissolves.

Notice that as the states empty out, both arrows point the same way: leaving speeds up and returning slows down at the same time. That's why the divergence in rates becomes very sharp.

Between these two states there must be a point where both events proceed at exactly the same rate. This is the equilibrium point of the pair "copper and its ion in this particular electrolyte." It's set by two properties: how tightly the copper lattice holds onto its atoms, and how readily the solvent accepts the resulting ion.

42.3. The numbers, and why they differ

Copper's equilibrium point sits around 3.4–3.5 volts on the lithium scale. Different papers cite values ranging from 3.1 to 3.7 volts.

The spread comes from two causes, and both are substantive. The first is electrolyte composition: the more readily the solvent stabilizes the copper ion, the more favorable it is for the ion to be in the liquid, and the lower the equilibrium point sits. The second is the steepness of the dependence: the rate grows exponentially with potential, by an order of magnitude for roughly every one hundred to one hundred fifty millivolts.

⚠ A correction on the steepness, added after checking. The first edition said "an order of magnitude every sixty millivolts" here. That's wrong. For a single-step, one-electron transfer with a symmetry coefficient of 0.5 at 25 °C, the theoretical slope is around 120 millivolts per decade; observed values range from 30 to 240 depending on the mechanism — 30–120 for activation-limited reactions and 120–240 for diffusion-limited ones. Sixty millivolts is a special case, not a general rule. For a specific reaction, the slope is determined experimentally.

So at 3.3 volts copper also dissolves, just so slowly that it goes unnoticed over the cell's life, while at 3.5 it happens visibly. There's no threshold in nature; there's a dependence so steep that the transition from "unnoticeable" to "catastrophic" fits within a few tenths of a volt. Every researcher has drawn the line wherever their instrument let them see it.

42.4. Why the graphite's potential decides copper's fate

The copper foil and the graphite mass are pressed against each other and electrically united — that's exactly what a current collector's job is. So they share the same electron states. The foil can't have a fill level of its own.

The graphite is what sets this level: there's incomparably more of it in the electrode, and it's exactly where the lithium and its electrons sit. The foil takes on whatever level the graphite has established — not by its own choice, but because it's bonded to it.

It's important here to keep two things apart that are easy to blur together.

Copper doesn't inherit any properties from the graphite: its own threshold stays its own. It only shares the current level with the graphite. Dissolution begins the moment the first quantity passes the second.

An analogy: the water level in a reservoir is set by the dam, while the mark below which a particular rock on the bottom becomes exposed is a property of the rock. The rock doesn't inherit the dam's properties; it's simply that when the level drops below its mark, it's exposed.

42.5. How this looks in operation

During charging, lithium enters the graphite together with its electrons, filling the states to capacity. The anode's potential of 0.05 volts is exactly what "overfilled" means. Copper is protected in this state: there's nowhere to place its electrons. In corrosion-protection engineering, this state is called cathodic protection; pipelines are protected this way deliberately, by connecting an external source, while the copper foil in a cell is protected unintentionally, by the mere fact of contact with charged graphite.

How seriously this is taken in manufacturing can be seen from the process: right after the electrolyte is poured in, the cell is given a partial charge to about 1.5 volts, to bring a bit of lithium into the graphite and pull the anode's potential below copper's corrosion threshold — and this is done even before the cell is left to soak. Copper is protected before the cell has even started working.

During discharge, lithium leaves and carries its electrons away. The states empty out, and the anode's potential rises. Within the normal range it reaches about one and a half volts — more than twice below copper's threshold. The potential can only reach the dangerous region under deep over-discharge.

The protection isn't lifted from outside. It was provided by the lithium in the graphite, and it ends together with it.

42.6. The numerical threshold at the terminals

It's easy to get this wrong, and the first edition of this material did get it wrong, taking the value "below 1.5 volts at the terminals" from the source material. That's incorrect. By measurement: on discharge from 2.5 to 2.0 volts, the anode's potential is 0.8–1.0 volts; at 1.5–1.8 volts at the terminals it only rises to 1.2–1.5 volts. Copper dissolution is observed at a terminal voltage of around 0.5 volts and below, and under forced over-discharge with polarity reversal down to minus 1.3–1.5 volts, the anode's potential reaches roughly 4.8 volts.

Dissolved copper enters the electrolyte as ions and settles as metal wherever the potential is below its equilibrium point. Deposits are found simultaneously at the anode, at the cathode, and on the separator's surface. The last of these is the most dangerous: metal deposited inside the insulating film builds a conductive bridge straight through it. From there the story repeats what happens with lithium needles — an internal short circuit.

The difference from the lithium case lies in irreversibility. Deposited lithium partly returns to service; dissolved copper never returns, and the risk it creates stays forever. A cell that has been through deep over-discharge must be considered safety-compromised, even if it shows normal capacity after recharging.

This is exactly why a battery management system disconnects the load well before the dangerous region, usually around 2.5–3.0 volts per cell, and why storing a discharged battery for a long time is more dangerous than storing one at half charge: self-discharge can carry the cell somewhere it can never come back from.

42.7. How over-discharge happens in a real pack

Not through any fault of the owner's. The management system usually measures voltage not per cell but per group; cells within a series string can end up below zero volts while the whole pack discharges. The weakest cell empties first, while the rest keep driving current through it, reversing its polarity. The pack was being discharged — and one cell inside it got over-discharged. This is also why per-cell voltage measurement isn't a luxury but a safety requirement for an aging pack, where the spread between cells keeps growing.

42.8. Over-discharge and overcharge are different accidents

Over-discharge is discharge below the lower limit: the anode empties out, its potential rises, and copper comes under attack, along with the protective film, which starts decomposing and releasing gas once too much lithium has been extracted. Overcharge is charging above the upper limit, and there the damage is almost entirely on the cathode side; it's covered in the next part.

There's a neat symmetry with copper here, and it explains how both current collectors are built.

Copper is protected by a state. It holds up because it's kept at a low potential — its electrons have nowhere to go. The protection disappears once the lithium in the graphite runs out.

Aluminum is protected by a film. Its own potential doesn't help it: at four volts, by the arithmetic of filling, aluminum should dissolve. What saves it is an extremely thin, impermeable crust on the surface.

And here's the surprise. The natural crust — aluminum oxide — isn't the main one. Hydrogen fluoride, formed when the LiPF₆ salt is hydrolyzed by trace water, converts the oxide into aluminum fluoride, and it's the fluoride that turns out to be the working protective layer.

In other words, hydrogen fluoride, which damages everything else in the cell, is the one thing keeping the aluminum foil from dissolving.

This explains something that would otherwise look absurd: why the industry sticks with the LiPF₆ salt despite its hydrolysis by water and the aggressive hydrogen fluoride it produces. Because in electrolytes using fluorine-free salts, aluminum undergoes extensive anodic dissolution. Replacing the salt destroys the foil.

When the protection fails. The film's stability depends on voltage nonlinearly: it peaks at around four volts, and above that, localized corrosion — pitting — begins. The reason is that the fluoride layer is uneven and folded, and local breakdown starts at its defects. Protons released by solvent oxidation also play a role: they etch away the oxide, after which the aluminum itself starts to oxidize. Elevated temperature worsens the film's ability to heal after local damage.

What this means in practice. The upper limit of 4.2 volts is constrained not only by the cathode's endurance but by the foil beneath it as well — and its point of best protection sits around four volts, meaning the top two tenths of a volt already operate beyond its best-protected zone. Pitting creates local punctures: at those spots the active mass loses contact with the current collector, and resistance rises unevenly across the electrode's area.

What matters here for comparing chemistries. Both current collectors are chosen for the potentials of a specific pair of electrodes. When the chemistry changes, the choice gets reconsidered: with LFP, the cathode operates at 3.2 volts, and the conditions for aluminum are gentler; in sodium-ion cells, sodium doesn't form an alloy with aluminum, so aluminum is used on both sides, and copper disappears from the design along with all of its problems. This is exactly why the breakdown is organized by component rather than by electrode.

A strip of the foil is left uncoated by the paste, and a metal tab — the current tab — is ultrasonically welded to it. Ultrasound is chosen because it doesn't melt the metal and doesn't damage the thin foil.

This is the one place in the cell where the joint is purely mechanical, and so the one thing that degrades from vibration.

What happens to it. Weld fatigue from vibration isn't an abstraction for automotive use — it's an operating condition. Fatigue from temperature cycling: the weld metal and the foil expand by different amounts, and every heat-cool cycle loads the boundary. Corrosion in the weld zone, where the metal's structure is disrupted and its protective properties are worse than in the intact foil.

How it shows up. Rising resistance at the junction — a contribution to the ohmic component, the one that sags instantly under current. Local heating at a point of poor contact, which speeds up everything around it. In the limit, a break — that is, complete cell failure, with no warning.

Why this is hard to notice. A weld defect produces a rise in resistance indistinguishable, by external measurement, from a resistance rise caused by anything else — pore clogging, drying-out, rearrangement of the cathode surface. The only way to tell them apart is by their characteristic timescale: the ohmic component sags instantly, the others over seconds and minutes (see the section on the three components of resistance in the notes on cell construction).

Three designs, with different vulnerabilities.

Cylindrical. A steel can with a crimped lid. The sturdiest, it withstands internal pressure and carries safety devices (section 47). The seal is a polymer gasket between the lid and the can; it ages, loses elasticity, and can allow a slow exchange with the atmosphere through it.

Prismatic. An aluminum casing with laser welding. Sturdy, but sensitive to swelling: gas pushes on the flat walls, and they bulge, disrupting the compression of the stack inside.

Soft, pouch type. Multilayer foil with a polymer coating, its edges heat-sealed. Lighter and cheaper, but the seal is the weakest point: it also serves as an emergency vent, opening when pressure rises.

What a loss of seal integrity produces. It works both ways. Electrolyte leaves — straightforward drying-out with all its consequences (section 31). Moisture comes in — a direct feed for salt hydrolysis and hydrogen fluoride production (section 29). Both processes speed up everything else, and both are irreversible.

Causes. Aging of polymer seals; corrosion of the casing from outside, including from electrolyte that has escaped through a micro-leak; mechanical damage; and swelling that stretches the seal on soft cells.

An external sign. Swelling is the only cell damage visible without instruments. As noted in section 30, it doesn't mean a standalone failure — it means that part of the electrolyte has turned into gas.

Components that exist only for the case of an accident. They're usually left out of a summary table of the cell's construction, because they play no part in normal operation.

The safety vent. A weakened section of the casing, or a separate membrane, that opens at a set pressure. It releases gas without letting the casing burst. On soft cells, the seal itself plays this role.

A pressure-triggered circuit interrupter. In cylindrical cells: a membrane that, as pressure rises, mechanically breaks an internal connection and cuts power to the cell.

A thermally sensitive element. A material whose resistance rises sharply on heating, limiting the current. It works against an external short circuit and, unlike the previous device, is reversible once things cool down.

What matters for this work. Only cylindrical cells have the full set. Prismatic ones usually have the vent, and not always the interrupter. Soft cells have only the seal opening. In other words, the level of internal protection is determined by the packaging, not the chemistry, and cells of identical composition in different packaging carry a different set of safeguards.

How these devices degrade. The circuit interrupter can trigger prematurely — from gas that has built up through ordinary aging rather than an accident; to the owner this looks like a healthy cell suddenly failing. The vent can become clogged with decomposition products and fail to open in time. The thermally sensitive element gradually loses its characteristic from repeated triggering and heating.

And the key limitation. All three devices protect against excess pressure and an external short circuit. None protects against an internal short circuit: once a bridge has already formed inside the stack, there's nothing left to disconnect — the current flows inside the cell, bypassing all of its terminals. This is exactly why puncturing the separator was singled out in the previous part as the one failure that leads to an accident rather than a loss of capacity.

What all of this part has in common. None of the components listed take capacity away gradually. They either work or they fail. That's why they play no part at all in state-of-health estimation — their condition isn't measured by instruments and can't be reconstructed from readings.

Thermal runaway has come up three times already — with lithium deposition, with oxygen release, with separator failure. Here it's treated as a process in its own right.

The essence is self-sustainment: each reaction releases heat, the heat triggers the next one, the next one releases more heat. Once the rate of heat release exceeds the rate of heat removal, the process stops depending on any external cause.

The order of stages, with temperatures. The values depend on chemistry, electrolyte composition, construction, and state of charge; for a given cell they're determined by testing. Below are the confirmed ranges.

75–100 °C: decomposition of the anode's protective film. It's the least stable thing in the whole cell. The reaction is exothermic; once it starts, bare graphite carrying stored lithium ends up in direct contact with the electrolyte.

Around 120 °C: the anode's lithium reacts with the electrolyte. Heat and flammable gases are released — ethane, methane, ethylene — and pressure rises.

130 °C for polyethylene, 170 °C for polypropylene: the separator's pores shut down. The one point where the process can still be stopped. If enough heat has already built up, it doesn't stop.

130–190 °C: the separator loses its shape and tears. The electrodes touch directly, adding the heat of a short circuit.

150–250 °C: cathode decomposition with oxygen release (section 20). The main stage: oxygen meets the organic electrolyte, and combustion runs inside the cell that needs no outside air. Onset of runaway for NMC occurs around 206 °C, for LFP around 244 °C, and LFP's olivine structure releases no oxygen at all under typical overload conditions — hence its fundamentally different level of safety.

Above 200 °C, the cathode's fluorine-containing binder joins the process, reacting with lithiated carbon and releasing heat.

Peak. The temperature at full runaway reaches on the order of 640 °C.

Dependence on charge. The severity of runaway rises above a state of charge of 50 percent; at full charge, the energy released can reach one to one and a half times the energy put in during charging. In other words, a fully charged battery isn't just a more likely case — it's also a more destructive one.

Why this can't be put out the ordinary way. The cell carries its own oxidizer. Cutting off the air supply, which works on an ordinary fire, is powerless here; the only thing that works is removing heat fast enough to break the self-sustaining loop.

Ways it can be triggered. An internal short circuit — from a lithium needle, copper deposits, or a particle that has punctured the separator. An external short circuit. Overcharge, where the cathode gives up oxygen while the anode gets coated in metallic lithium. Mechanical damage. External heating, including heat from a neighboring cell. And its own heating under load once resistance has grown.

Delay. An internal short circuit doesn't have to happen at the moment of the damaging event. A needle grows episode by episode, and the puncture occurs days or weeks after the last cold fast charge — at rest, with no external trigger. This is the one process in the cell where damage to service life and damage to safety are separated in time.

Why an older cell triggers more easily. Every wear pathway already covered lowers the threshold. Particle cracking has produced more reactive surface. Deposited metallic lithium is itself extremely reactive. The separator is thinned and partly clogged, so current flows unevenly. There's less electrolyte, so heat dissipation inside is worse. Resistance is higher, so under the same load the cell heats up more.

And there's a numerical measure for this. Lithium plating brought on by fast charging can lower the onset temperature for self-heating to 60 °C — well below the thresholds normally expected to cause trouble. In other words, a cell that has been through episodes of cold or excessively fast charging enters the danger zone at temperatures that are safe for a healthy one.

In other words, a cell's safety isn't a fixed quantity set by its design — it declines together with its remaining service life. This is a separate argument for why state-of-health estimation matters for more than just residual value.

This brings together what was covered in section 34 and in the sections on lithium deposition.

Cold slows all chemistry down by roughly half for every ten degrees. The difference between minus twenty and plus twenty-five is around a twentyfold slowdown. A year in the cold ages a cell about as much as two to three weeks in the warmth. It slows things down, but doesn't stop them.

Cold by itself does harm in two ways with no current involved at all: possible precipitation of solvent components, and uneven thermal contraction of the materials, which stresses the boundaries between them over repeated trips through freezing temperatures.

The main danger arises only in combination with charging current. Discharging in the cold is harmless: lithium moves into the cathode, where it has nowhere to be metallic, and all that happens is a voltage sag. Charging is dangerous: lithium moves into the graphite, all three steps of its entry are slowed, and if it's delivered too fast, it settles as metal on the outside.

How this differs from heat's harm. Heat speeds up what happens anyway. Cold with charging current triggers a mechanism that doesn't exist at all in the warmth. So the phrase "cold speeds up degradation" is imprecise: cold opens an additional path to damage while simultaneously slowing down all the others.

Everything covered above has been processes. Instruments can't see processes: all they can see is consequences. There's no one-to-one correspondence between the two, and that's the main difficulty in state-of-health estimation.

A reminder of the four consequences from Part I. Three of them are called degradation modes, because each distorts the open-circuit voltage curve in its own way: lithium dropped out of circulation, sites lost in the anode, sites lost in the cathode. The fourth stands apart and doesn't count as a mode: increased resistance, which shows up not on the rest-voltage curve but as a sag under load.

The first thing the table shows. The last column is filled in completely. Every process produces rising resistance, without exception — and so, on its own, it points to nothing specific. Its diagnostic value is zero unless its components are separated by their characteristic timescales.

The second thing. Only three rows produce exactly one consequence: separator clogging, weld defects, and metal deposition. The rest produce two or three.

Can be told apart: lithium loss versus site loss. These are different distortions of the voltage curve. With lithium loss, the electrodes stay intact, and the curve shifts as a whole. With site loss, the balance between the two electrodes' capacities changes, and the curve deforms — one electrode's working point shifts relative to the other's.

Can be told apart: anode versus cathode. Site loss in the anode and in the cathode shifts the curve in different directions, and on the curve's derivatives this shows up as a divergence in characteristic points.

Blurs together: causes within one consequence. Sites in the anode are lost through five paths — cracking, binder failure, graphite exfoliation, drying-out, foil corrosion. All five produce the same shift. No voltage measurement will tell them apart.

Blurs together: rising resistance from different causes. Separating by timescale helps partially: the instantaneous component points to the foil, the welds, and the electrolyte volume; the medium one points to interfaces and surface layers; the slow one points to transport inside particles and drying-out. But within each component, the causes remain indistinguishable.

Blurs together completely: resistance versus capacity when measured under load. A cell with increased resistance hits the lower threshold before it's actually empty, and the instrument records a drop in capacity that doesn't physically exist. The only way to tell them apart is by measuring at different currents: for a cell that has lost lithium, the result barely depends on current; for a cell with increased resistance, it depends heavily.

This is the most important line in the entire part.

Among everything covered, none of the following give any warning sign: a dendrite puncturing the separator, a particle pressing through it, thermal shrinkage, vent failure, premature triggering of the circuit interrupter, weld fatigue up until the moment it breaks, and accumulated risk from past episodes of cold charging and over-discharge.

All of these belong to safety, not to capacity. A cell with a lithium needle growing inside it shows normal capacity, resistance, and voltage curve — right up until the moment of the puncture.

This leads to a conclusion that shapes the architecture of any state-of-health estimate: measured state and safety are not the same thing. A health estimate based on capacity and resistance answers the question of residual value and remaining range. It doesn't answer the question of risk, and it can't be extended to answer it — that requires usage history: whether there were episodes of cold charging, deep discharge, overheating, or fast charging on a worn cell.

A section about the physics of the signature, not about instruments: exactly how the three modes show up in the shape of the open-circuit voltage curve. Extracting this signature is the subject of a separate document on measurement.

The starting point. A cell's voltage curve is the difference between two curves: the cathode's potential and the anode's potential, each depending on how full it is with lithium. As long as both curves are in place and aligned the way they were at manufacture, the cell behaves normally.

Lithium loss. The electrodes are intact, and neither of their curves has changed. One thing has changed: there's less lithium, and it no longer fills the anode to the same degree as before. The electrode curves shift relative to each other, keeping their shape. The result: capacity has dropped, the cell's curve shape is preserved, and its characteristic features have shifted along the charge axis without changing their appearance.

Loss of sites in the anode. The anode's curve compresses along the charge axis: the same potential range fits into fewer amp-hours. The cathode's curve doesn't change. The difference between the two curves deforms: features on the anode side move closer together, while those on the cathode side stay at the same distances.

Loss of sites in the cathode. A mirror image: the cathode's curve compresses, the anode's stays intact.

This is where the distinguishability comes from. Lithium loss produces a shift with no deformation; site loss produces deformation. And anode loss differs from cathode loss in which part of the curve compressed. This is exactly the basis for telling the three modes apart from voltage alone — provided the curve is recorded in enough detail and at rest.

What this signature doesn't contain. The causes within each mode, as noted in section 55. The curve says the anode has fewer sites, and it doesn't say whether that's from cracking, flaking, or drying-out.

Three conclusions that the next part is built on.

Resistance by itself is useless; split by timescale, it's useful. So measurements that give only a total resistance provide less than they appear to.

Capacity measured at a single current is ambiguous. So data at different currents is needed, or an explicit separation of the resistance contribution.

Usage history carries information that current measurements don't contain. So it isn't supplementary information — it's an independent source, without which part of the picture is fundamentally out of reach.

Everything covered up to now described a cell running its own affairs. In a pack, that stops.

A cell no longer chooses its own current — its neighbors set it. It doesn't choose its own temperature — its position within the volume determines it. It doesn't choose when to stop discharging — the controller decides that for the group as a whole. And its own state stops being its own business: the weakest cell limits all the others.

This gives the rule underlying the whole part: a cell's degradation processes inside a pack are the same, but the conditions driving them are set from outside. On top of that, mechanisms appear that a standalone cell simply doesn't have.

A typical traction pack has around a hundred cells in series, sometimes with two or three in parallel at each position. The two connection types behave in opposite ways.

Series. The same current flows through all of them, and the voltages add up. A cell can't take less current than flows through the string, even if that's hard on it. Divergence between cells is visible here: each has its own voltage, and the controller measures it.

Parallel. The voltage is shared, and the current splits on its own, inversely proportional to resistance. There's a weak self-balancing effect here: a cell with higher resistance automatically takes less current. The price for this is invisibility: a parallel group has one voltage for all its cells, and the controller measures the group, not the individual cells within it.

What this means in practice. In a pack with a hundred positions of two cells each, the controller sees a hundred voltages, not two hundred. The failure of one cell inside a parallel pair doesn't show up directly in the measurements.

Failures there also differ in their consequences. If a cell goes open-circuit, its partner takes double the current and ages twice as fast, dragging the whole position down with it. If a cell goes to a short circuit, it discharges its partner through itself, and the whole position sags.

A consequence for this work. The depth of per-cell observation is limited not only by data access but by the design itself: parallel groups are fundamentally opaque. The wiring scheme sets a resolution limit, and it needs to be known before assessing repairability.

In a series string, charging stops when the first cell reaches the upper limit, and discharging stops when the first one reaches the lower limit. Not the average, not the majority — the first one.

A pack's usable capacity is set by its worst cell, not its average one. Ninety-five cells at 90% capacity and one at 60% give a pack that behaves like a sixty-percent one.

An economic consequence. A pack showing 70% health could be two very different assets: ninety-six evenly worn cells — or ninety-five good ones and one bad one. In the first case, there's nothing to recover. In the second, replacing a single module brings the pack back to ninety percent.

These cases can't be told apart from a single health number. Only the spread tells them apart.

Three sources, and the third is qualitatively different from the first two.

Initial manufacturing spread. Nominally identical cells differ in capacity, resistance, and self-discharge rate. Manufacturers sort them before assembly, but a perfect match never happens.

Systematic spread by position. A cell in the middle of the volume dissipates heat worse than one at the edge and runs hotter. By the law of the rate doubling for every ten degrees, a ten-degree difference means twice as fast aging. This isn't random — it's a pattern: for a given design, the map of fast-aging positions stays fixed.

Self-reinforcement. A cell with higher resistance heats up more under the same load; the heat speeds up its degradation; the degradation raises its resistance further. The loop closes.

This means the spread doesn't just accumulate over time — it accelerates. A pack starts life nearly uniform, and the non-uniformity grows faster than the average wear. This is exactly what makes spread a more sensitive indicator than average capacity: it reacts sooner.

This also explains the end-of-life bend in a pack: it comes not only from electrolyte drying-out inside the cells (section 31), but also from the spread reaching a magnitude at which the worst cell sharply limits all the others.

The controller balances the cells. Usually passively: excess charge from the fuller cells is bled off through resistors as heat. Less often actively, by shuttling charge between cells.

It can equalize state of charge — bring every cell to the same fill level.

It cannot equalize capacity and resistance. A cell at 60% capacity, after perfect balancing, is still a cell at 60% capacity.

What follows from this. Balancing masks the early stages of divergence: as long as its capacity is enough, the spread in voltages looks small, even though the cells have already diverged in capacity. By the time the spread becomes visible in the voltages, the divergence is already significant.

A diagnostic conclusion. Balancing activity itself is a signal. A cell that has to be trimmed constantly, and always in the same direction, has elevated self-discharge: a defect where capacity is normal, but the cell keeps ending up below its neighbors regardless. Data on balancing activity by position carries information that isn't present in the voltages themselves.

Within the pack. Central cells run hotter than edge ones; those closer to the cooling plate run cooler than those farther away; with air cooling, the first ones in the airflow run cooler than the last.

Within a single cell. The area around the current tabs heats up more than the middle — that's where all the current converges. So even a single cell ages unevenly across its volume.

There are fewer sensors than cells. Usually just a few per module. The pack's hottest spot typically isn't covered by a sensor, and its temperature is estimated by a model rather than measured.

A consequence. The thermal spread is systematic, not random, and so predictable: for a given design, you can say in advance which positions will age first. A rare case where degradation can be forecast from geometry rather than from measurement.

A mechanism that exists only at the pack level: a cell in runaway heats its neighbors, and they enter runaway in turn.

The speed and likelihood of propagation are set by the design: the spacing between cells, thermally insulating layers, the direction gases are vented, how well cooling works. This is a property of the pack, not of the chemistry — identical cells in different layouts give different outcomes.

What this means for the asset. The failure of a single cell can destroy the entire pack. The risk isn't proportional to the number of bad cells: it's determined by whether at least one exists and by the quality of the structural barriers. One more argument for why safety assessment and capacity assessment are different tasks.

The controller is usually thought of as a measuring device. But it's electronics sitting inside the same pack, under the same thermal conditions, and it ages.

Calibration drift. The accuracy of voltage measurement drifts over time. An error of a few millivolts per cell is noticeable in the charge estimate, since the working curve is flat in places.

Loss of reference points. The charge estimate is built by tallying charge passed, corrected against rest voltage. The correction requires the cell to occasionally sit near the edges of its range — that's where the curve is steeper and less ambiguous. A car that's always driven between thirty and seventy percent never gives the controller those points, and the accumulated counting error never gets reset.

This leads to something unexpected: gentle usage makes the state-of-health estimate worse, even though it improves the actual state. The reading can diverge from reality more for a carefully treated battery than for one used across its whole range.

The model drifting from reality. The controller's model describes a cell of a particular age and chemistry. As the cell wears, it moves away from the model, and the estimate gets worse exactly when it's needed most.

Sensor and communication failures. A failed temperature sensor leads either to conservative power limiting or to operating blind. Communication faults between measurement boards cause parts of the data to drop out.

Cells are connected by busbars, modules by power cables and connectors.

What happens. Bolted joints loosen from vibration and temperature cycling. Where dissimilar metals meet — an aluminum busbar to a copper terminal — galvanic corrosion sets in. Contact resistance rises, local heating rises, and the heat speeds up the corrosion.

Why this matters diagnostically. Rising connection resistance looks, in measurements, exactly like cell degradation: the pack sags harder under load, and this symptom alone can't tell you whether the cells have aged or a contact has loosened.

How to tell them apart. Per-cell voltages are taken at the cells' own terminals, while pack voltage is taken at the pack's terminals. The difference between the sum of the per-cell voltages and the pack voltage under load is the drop across everything in between: the busbars, joints, and connectors. In a healthy pack this is small and stable; a rise in it points to the connections, not to the chemistry.

This is the most accessible system-level diagnostic there is: it needs the same data as everything else, and nothing more.

What degrades. The pump wears out; the coolant loses its properties and gets contaminated; channels clog; seals leak. With air cooling, filters clog and fans fail.

Why it's dangerous. Not through failure, but through silence. Weakened cooling produces no error and no symptom — it simply raises the operating temperature by a few degrees. By the law of doubling every ten degrees, this means accelerated aging of the whole pack that looks like natural wear.

A leak inward. Coolant inside the pack means dropping insulation resistance and short-circuit risk. The one system failure detected directly: insulation is monitored continuously and reacts immediately.

Sealing. The pack is protected against water and dust, and that protection degrades: seals age, mounts wear from vibration, and mechanical damage is possible.

Corrosion from outside. For northern climates this isn't a background detail: road de-icing chemicals, prolonged contact with salt solution, freeze-thaw cycles. The underside of the pack is the most heavily loaded surface in this respect.

Mechanical damage. An impact against an obstacle from below may not breach the casing, but it can deform it and disrupt the stack's compression or the geometry of the cooling channels.

The pressure-equalization membrane. It lets air through and keeps water out; it gets contaminated and loses its properties, at which point the pressure difference loads the seals.

Insulation monitoring is the one parameter in this group measured directly and continuously. It reacts to moisture, to a coolant leak, and to damaged cable insulation alike. In effect, it's the only direct measure of the pack's health as a structure.

Three conclusions that shape the approach taken here.

A pack's health isn't the average health of its cells — it's a function of the worst one. An average figure with no spread hides exactly what determines residual value.

Several system-level failures — connections, cooling, the controller — look like chemical degradation. They can be told apart, but only with per-cell data compared against the pack-level totals.

Spread reacts sooner than average capacity does, because it accelerates on its own. So the early sign of a pack deteriorating isn't a drop in capacity — it's rising non-uniformity.

A bringing-together of what's scattered across Parts II–VIII into one picture. The distinction is practical: one battery sits idle, another works, and they don't just wear at different speeds — they wear differently.

Calendar aging runs on the clock. The battery sits unused, and the wear continues.

Cyclic aging runs on the number and depth of cycles. When the battery sits idle, these processes don't happen at all.

Three things that follow from this.

Their shape over time is different. Calendar aging runs roughly as the square root of time: fast at first, then slower and slower (section 5). Cyclic aging is closer to linear in the number of cycles — up to the bend, after which it collapses.

They don't just add together — they feed each other. Cracking exposes fresh surface, on which the calendar-driven film growth speeds up. That's why the square-root law doesn't hold for a battery that's being worked.

They're driven by different levers. Calendar aging is driven by temperature and state of charge during storage. Cyclic aging is driven by cycle swing and current. A battery that's driven little but sits fully charged in the heat can age faster than one worked hard under moderate conditions.

How to read this table. The rows aren't equal in strength. Temperature acts on everything, and so it's the only factor that harms both electrodes equally. State of charge acts selectively, but at the cathode it drives eight processes out of ten. Cycle swing doesn't touch chemistry at all, yet it alone is enough to mechanically destroy the electrodes.

What this yields as an operating rule. Three levers, each drawn from the table rather than from general considerations: don't hold a high charge without need; don't keep it in the heat; don't charge it fast in the cold. The first two extend service life, the third is a matter of safety.

The check was carried out on 01.08.2026 against the load-bearing claims — the ones the document's conclusions rest on. Result: two confirmed verbatim, two confirmed with a divergence, three not confirmed, "source not about that" — zero. Not a single source turned out to be closed to machine reading.

The slope of the rate-versus-potential dependence. Was: an order of magnitude every 60 millivolts. Found: for single-step, one-electron transfer with a symmetry coefficient of 0.5 at 25 °C, the theoretical slope is around 120 millivolts per decade; observed values are 30–240 depending on the mechanism. Consequence: the estimate of the acceleration between 4.0 and 4.2 volts was corrected from a thousandfold to roughly fiftyfold, and the combined figure with heating from four thousandfold to roughly two hundredfold. Corrected in sections 28.1, 28.2, 42.3.

The threshold for oxygen release from NMC. Was: once more than 50–60 percent of the lithium is extracted. Found: release begins at roughly 80–81 percent extraction, and it's governed by the fraction of lithium removed, not by potential. Onset for NMC811 is around 4.3 V versus lithium (≈4.2 V at the terminals), for NMC111 and 622 around 4.7 V (≈4.6 V at the terminals). The figure of 50–60 came from the source material. Corrected in section 20.

The reason for using rubber-based binder on the anode. Was: the fluorine-containing polymer gets reduced there. Found: fluorine-containing binder is used on anodes and works there; the reasons for choosing rubber-based binder are cost, adhesion, water-based processing, and a smaller irreversible first-cycle loss. Corrected in section 13.

The sequence of thermal runaway. The order of stages is correct, but the temperatures weren't given. Confirmed ranges have been added, along with the dependence of severity on state of charge. Section 49.

The thickness of the rock-salt layer. Was: a few nanometers. Found: up to 100 nanometers after 500 cycles at 80 °C. Corrected in section 23.

Ionic radii of Li⁺ at 0.76 Å and Ni²⁺ at 0.69 Å in octahedral coordination; increased cation mixing in high-nickel compositions.

Volume change of the layered structure under two percent on lithium insertion and extraction.

A drop in the onset temperature for self-heating to 60 °C from lithium plating — section 50.

The dependence of runaway severity on state of charge, with a threshold of 50 percent — section 49.

The exothermic reaction of the fluorine-containing binder with lithiated carbon above 200 °C — sections 13 and 49.

Electrolyte composition in percentage shares (60–80 solvent, 10–15 salt, 1–5 additives) — taken from the source material, not cross-checked against primary sources.

Separator pore size of 20–50 nanometers — same status.

The anode-to-cathode capacity ratio of around 1.1 — an industry-wide rule of thumb, not checked for any specific cell.

Whether direct recycling of cathode material is applicable under local conditions — not checked.