The sizing article treats degradation as a curve you plan around. This article opens the hood: why does the curve exist, what physically eats capacity, and which operating choices feed which mechanism. Once you know the culprits, every warranty condition — temperature caps, SoC limits, C-rate ceilings — stops looking arbitrary and starts looking like a rap sheet.
For a visual companion, explore the interactive Lithium-Ion Cell diagram on BESS.Engineer.
The shape of the fade: three stages
Before the mechanisms, the shape. Nearly every warranty curve tells the same three-act story — and knowing the acts tells you which mechanism is on stage.
Stage one — the initial drop. In the opening months a cell sheds its first few percent. This is the SEI film finishing the settling it began during factory formation, plus a little early lithium bookkeeping. On a brand-new asset it can look alarming. It is entirely normal, and good warranties price it in.
Stage two — the long linear middle. For most of the project life the curve is a gentle, near-straight decline. This is the steady grind: calendar-driven SEI growth ticking away with temperature and state of charge, cyclic fatigue scaling with throughput. Nothing dramatic happens here, which is exactly the point. A well-sized, well-cooled LFP system spends fifteen-plus years on this stretch, losing well under a percent of capacity a year.
Stage three — the knee. Late in life the slope can bend sharply downward as mechanisms begin feeding each other; the knee section below has that physics. End of life is a contractual line drawn before the knee: commonly 65–70% of beginning-of-life capacity, and OEM-specific, so the asset is retired while still on the predictable part of the curve. Sizing and augmentation (adding capacity over the project’s life to offset fade) exist to keep a project on stages one and two, and off stage three.
Mechanism 1: SEI growth — the slow lithium tax
The SEI (solid-electrolyte interphase) is the protective film on the anode, born during the factory’s formation cycles. It never fully stops growing: tiny ongoing reactions keep thickening it throughout life, and every bit of growth is built from lithium withdrawn permanently from the working inventory. This is the steady, patient mechanism behind the long near-linear middle of the fade curve, and behind calendar aging: the fact that a battery fades even sitting unused. Its accelerants are heat (reaction rates roughly double for every ~10 °C of cell temperature) and high state of charge (a fuller anode is a more reactive one). That pairing, hot and full, is what warranty SoC-dwell and temperature clauses exist to police, and what makes pre-lithiation valuable: extra inventory to feed the tax.
Mechanism 2: Lithium plating — the dangerous shortcut
During charging, lithium ions are supposed to slot into the graphite. Push them faster than the graphite can accept — by charging at high C-rates, at low temperatures where solid-state diffusion is sluggish, or at very high SoC where the graphite is nearly full — and ions give up queueing and deposit as metallic lithium on the anode surface. Plating is degradation’s villain twice over: the plated lithium is largely lost to capacity, and the metallic deposits can grow dendrites, needles that threaten the separator and raise internal-short risk. This is why charging below 0 °C is restricted in graphite-anode lithium cells, LFP included, and why cold-climate designs heat their batteries before charging. Fast charging, cold charging, charging near full — each invites plating, and combinations compound the invitation.
Mechanism 3: Mechanical fatigue — particles that crack
Every cycle, electrode particles swell and shrink as lithium moves in and out. Over thousands of repetitions this breathing cracks particles, fractures conductive pathways, and exposes fresh surfaces that promptly grow new SEI (paying the lithium tax again). This is the heart of cyclic aging, throughput-driven wear, and it scales with depth of discharge and rate: bigger swings, bigger breaths, faster fatigue. It’s also why chemistries differ so much in cycle life. LFP’s small volume change and its more stable cathode structure are the core reasons it out-cycles NMC, often past 6,000 full cycles to 80%.
Supporting cast
Two more names you’ll meet: electrolyte decomposition (the liquid itself slowly degrades, especially hot, thickening interfaces and raising resistance) and, in nickel-rich cathodes, transition-metal dissolution and structural fatigue that migrate to poison the anode. Resistance growth deserves its own respect: a cell can fade in power faster than in energy, sagging under load while still holding decent charge. End of life, though, is almost always written against energy capacity, which is why a battery can still pass its warranty on paper while quietly losing the power to hold its dispatch curve.
The knee — when mechanisms gang up
The knee is not a new mechanism; it is the old ones starting to feed each other. Plating roughens the anode; rough surfaces grow SEI faster; thicker SEI raises resistance and local heating; heat accelerates everything. Once those loops close they don’t reopen, which is why the commercial rule is blunt: operate inside the envelope so they never close in the first place. Projects that plan to visit the knee don’t get to plan much else.
The duty writes the curve
Here is the part that surprises people new to storage procurement: the biggest driver of how fast a battery fades usually isn’t the battery. Hand the same cell to five different projects and you get five different curves.
Illustrative. The real ranking shifts with state-of-charge dwell, depth of discharge, C-rate and temperature — a hard-cycled merchant asset can fade faster than a reserve unit, or slower, depending on all four.
A peak-shaver that discharges gently once a day, kept cool and rested at a moderate state of charge, barely leaves the linear middle in twenty years. A frequency-regulation asset micro-cycling thousands of times a day, or a merchant battery worked hard on every price spread, drives heat and throughput far faster and reaches its end-of-life line years sooner. Same chemistry, same enclosure, different duty.
That is why a warranty is never a single number. It is a number conditional on an envelope: cycles per year, maximum depth of discharge, temperature limits, SoC-dwell restrictions, C-rate ceilings. Each clause is the vendor pricing one of the accelerants from the mechanisms above. Push past the envelope and the curve bends toward the knee, warranty or no warranty.
The operator’s cheat sheet
Heat feeds SEI growth; cold-plus-charging feeds plating; depth and rate feed fatigue; sitting full feeds calendar aging. The countermoves fall out of the same list: hold a middle state-of-charge band, cycle shallower whenever the market lets you, keep the cells cool, and never fast-charge them cold. Every line in a warranty envelope is one of those sentences turned into a number.
FAQ
Which mechanism dominates in grid storage? Gentle duty (low C-rates, controlled temperature) makes calendar-driven SEI growth the usual lead, with cyclic fatigue scaling alongside throughput. Abuse changes the ranking fast.
Can degradation be reversed? The mechanisms above are one-way. Some apparent recoveries exist (rest, rebalancing recovering measured capacity), but lost lithium and cracked particles don’t come back — plan on management, not miracles.
Why do warranties ban charging below freezing? Lithium plating. Cold graphite accepts ions too slowly, so charge current deposits metal instead — capacity loss now, safety risk later. Heaters exist for exactly this.
The degradation chapter of my Grid-Scale BESS: Complete Guide connects these mechanisms to warranty tables, dispatch strategy, and augmentation math.