Lithium-Ion Cell Manufacturing: Slurry, Slot-Die Coating, Calendering and Formation
A lithium-ion cell factory is not a chemistry plant. It is a roll-to-roll coating line that happens to end in an electrochemical device. The cathode active material arrives as a powder with a datasheet specific capacity, goes through a mixer, a slot-die head, a 90-metre dryer, a two-roll calender, a slitter, a stacker and three weeks of formation — and every one of those steps can destroy 5% of the value that has accumulated before it.
The economics are brutal in a specific way: materials are 78–82% of cell cost for mainstream chemistries, and the cathode alone is roughly 53% of cell material cost. A cell scrapped after formation carries the full cost of everything already done to it. At \74/kWh average cell price (BloombergNEF 2025 survey), a 5% scrap rate on a 10 GWh line throws away about \37 million a year of material that was already converted into something almost, but not quite, a battery.
This is the process engineering behind that number: what the coating weight tolerance actually does to porosity, why the anode is deliberately oversized, why drying fast is cheap and dangerous, and why the last three weeks of the line are the most capital-intensive real estate in the building.
1. The line, end to end
flowchart TD
A[Active material<br/>CAM / graphite] --> B[Mixing<br/>planetary + high-shear]
B --> C[Slot-die coating<br/>wet 150-250 um]
C --> D[Multi-zone drying<br/>NMP recovery or water]
D --> E[Calendering<br/>30-40% porosity]
E --> F[Slitting<br/>burr under 15 um]
F --> G[Winding or stacking<br/>dry room -40 C dp]
G --> H[Tab welding<br/>ultrasonic / laser]
H --> I[Electrolyte filling<br/>vacuum, -60 C dp]
I --> J[Wetting 6-24 h]
J --> K[Formation 0.1-0.5C<br/>SEI]
K --> L[Degassing and reseal]
L --> M[Aging 1-3 weeks<br/>K-value sort]
M --> N[Grading and shipment]
Two structural facts follow from this flow:
The line is asymmetric in time. Coating runs at 30–80 m/min. Formation and aging run at days-to-weeks. That mismatch is why a cell plant's floor space is dominated not by the coaters but by racks of cyclers and environmental chambers — aging alone can occupy up to a quarter of the plant footprint.
The line is asymmetric in value. A defect introduced at mixing costs the slurry. The same defect detected at end-of-line costs the slurry, the foil, the separator, the electrolyte, the formation energy, the dry-room time and the depreciation on a very expensive building.
2. Mixing: the rheology that decides everything downstream
A cathode slurry is a concentrated suspension: 94–96 wt% active material, 2–3% conductive carbon (typically a chain-structured carbon black at 40–60 m²/g), 2–3% PVDF binder in NMP, at 68–76% solids. It behaves as a shear-thinning, yield-stress fluid, and the slot-die head only works if that rheology is stable over the hours between mixing and coating.
The relevant model is the power law:
with a yield stress \tau_0 of a few pascal, a consistency index K in the tens of Pa·s^n, and a flow index n around 0.3–0.6 for well-dispersed cathode slurries. Slot-die coating requires the slurry to be shear-thinning enough to pass the die slot at moderate pressure, but elastic enough not to drip or sag after deposition. The viscoelastic balance is set by the carbon black network, which is why carbon type and dispersion energy matter more than most process engineers expect: a poorly dispersed carbon black gives both high resistivity and a slurry that settles between the mixer and the die.
Anode slurries are a different fluid. Water-based graphite uses CMC as a thickener and SBR latex as the binder — a two-component system where the SBR is dispersed as 100–200 nm particles and the CMC adsorbs on graphite edges. Solids run 45–55%, viscosity is lower, and the drying behaviour is dominated by water's much higher latent heat and its tendency to cause binder migration to the coating surface.
Two contamination limits sit at the mixing stage and are worth stating because they are frequently violated in new plants:
- Magnetic contamination. Iron and stainless wear particles from mills, pumps and piping are the most common cause of self-discharge outliers. Specs are typically in the single-digit ppm for the finished electrode, with magnet traps at every transfer and ceramic-lined pump bodies.
- Moisture in the powder. NMC and LFP are hygroscopic; residual surface water reacts with PVDF and with the electrolyte salt. Active material is dried before mixing, and the mixer itself is often blanketed.
3. Coating: from grams per square metre to micrometres
Slot-die coating is a pre-metered process: the pump sets the wet film thickness, and the die gap sets uniformity. Comma-bar coating is the older alternative, metering by a blade gap; it is cheaper and more forgiving of viscosity but produces worse cross-web uniformity at high speed.
The first calculation every process engineer runs is the wet thickness required to hit a target dry loading:
where m_d is dry coating weight in g/m², \rho_{\text{wet}} is slurry density (1.35–1.55 g/cm³ for NMP cathodes) and s is the solids mass fraction.
For a 180 g/m² NMC622 cathode at 70% solids and 1.45 g/cm³:
A 177 µm wet film becomes roughly 85 µm dry and, after calendering, about 55–60 µm. That is the whole game: three orders of magnitude between the coating gap and the final tolerance, with the tolerance budget measured in tenths of a milligram.
And the tolerance budget is tight. Published manufacturing-tolerance analysis for commercial electrodes uses ±0.4 mg/cm² on coating weight and ±3.0 µm on calender thickness. Those numbers sound trivial. Their effect is not: at a nominal 2.79 mAh/cm² loading with a 30% target cathode porosity, the ±0.4 mg/cm² and ±3 µm bands produce a porosity range of 19.6% to 38.6%. Electrolyte volume required to fill the pores then ranges from 59% to 141% of the target fill. That single result explains most of the cell-to-cell variation that shows up as rate-capability spread in a shipment.
Convert loading to capacity with the active material's practical specific capacity:
where f_{\text{AM}} is the active-material mass fraction (0.94–0.96). For NMC622 at 180 mAh/g and 0.95 active fraction, 180 g/m² gives:
That sits in the middle of the commercial band. Published surveys of production cells cluster around 1.4, 2.8 and 4.2 mAh/cm² for high-power, mid-range and high-energy designs respectively. Above about 4.5 mAh/cm², electrolyte transport through a 30%-porosity electrode starts to dominate the rate capability, and above 5 mAh/cm² the coating also becomes mechanically difficult to dry without cracking.
Throughput arithmetic
Consider a 1.4 m wide coater at 40 m/min, double-sided, coating 3 mAh/cm² (30 Ah/m²):
At 3.65 V nominal, that is 0.736 MW of cell capacity per hour. A 10 GWh/year plant running 6,000 productive hours needs 456 kAh/h of cathode — so roughly 2.3 coater-passes at this speed and width, before OEE losses. This is why coater speed is the single most levered number in cell plant design and why the industry has pushed web speeds from 30 m/min for thick electrodes toward 60–100 m/min.
4. Drying: where the money and the defects both come from
Drying is the energy centre of the plant. NMP-based cathode drying ovens routinely run 50–100 m long and can consume up to half of a cell factory's energy, which is the entire economic argument behind solvent-free electrode processing.
The physics is a two-phase problem. In the first phase, solvent evaporates from a shrinking film at a near-constant rate controlled by the external mass transfer. In the second phase, the film has shrunk to a porous solid and solvent must travel by capillary action to the surface. Binder and carbon black travel with it. If the second phase is driven too hard, binder accumulates at the top surface and the electrode becomes binder-starved at the foil interface — the coating then delaminates during calendering or, worse, during cycling.
The controlling dimensionless group is a Peclet number for the binder:
When Pe \gg 1, the evaporation front outruns binder diffusion and the binder profile becomes non-uniform. Practical process windows follow directly: reducing the drying rate in the capillary phase from 1.19 to 0.52 g/m²/s has been shown to raise adhesive strength by 50%. Additive chemistry can partially rescue the fast-drying window — 0.5 wt% of a suitable additive has been shown to suppress binder migration up to 6 g/m²/s for a 4.2 mAh/cm² graphite anode, enabling drying times near 30 s.
Drying capacity is a length problem, not a temperature problem:
At 80 m/min with a single-stage profile, a 4.2 mAh/cm² anode needs about 92 m of dryer. Two-stage drying cuts it to 62 m; three-stage to 43 m. That 53% length reduction is a direct capital-cost reduction and, in a building where the dryer sits inside a controlled atmosphere, a direct energy reduction as well.
For NMP cathodes, the solvent cannot simply be vented. Recovery is by condensation or absorption, with recovery rates above 99% required for both economics and emissions compliance; the recovered NMP is distilled and returned to the mixer. Water-based anode drying avoids the recovery plant entirely but pays for it in evaporation energy: water's latent heat is roughly 2,260 kJ/kg against NMP's ~470 kJ/kg, so per kilogram of solvent the anode line needs nearly five times the heat — offset by the fact that anode coatings are thinner and run faster.
5. Calendering: buying energy density with rate capability
Calendering compresses the dried coating between two hardened rolls, reducing porosity and increasing volumetric energy density. It is the single most consequential trade in cell design, because porosity controls both how much active material fits in a litre and how fast lithium ions can move through the electrolyte-filled pores.
Porosity follows from coating weight and thickness:
For NMC622 (\rho_{\text{true}} \approx 4.46 g/cm³) at 180 g/m² and 85 µm as-dried:
Calendering to 58 µm gives:
That is a 1.47:1 compression ratio. Typical targets are 30% for nickel-rich cathodes, 25–30% for LFP, and 30–40% for graphite anodes — with the anode usually left more porous than the cathode because it must accept lithium faster than it releases it and because graphite particles are more prone to fracture under load.
Effective transport in the compressed electrode degrades with the Bruggeman relation:
At \varepsilon = 0.30, the factor is 0.164; at \varepsilon = 0.52 it is 0.375. Halving porosity cuts effective diffusivity by more than half, which is why a 4.2 mAh/cm² electrode calendered to 25% porosity charges noticeably slower than the same electrode at 35% — and why fast-charge cells deliberately give up 5–8% volumetric energy density for pore volume.
Two mechanical failure modes accompany over-calendering:
- Particle cracking. NMC secondary particles are polycrystalline agglomerates; above roughly 3.6–3.8 g/cm³ electrode density the primary grains crack, exposing fresh surface that consumes lithium in new SEI. Capacity fades faster even though the first-cycle capacity looks fine.
- Foil elongation and wrinkling. Aluminium foil at 12–15 µm and copper at 6–9 µm both elongate under calender load. Differential elongation between coating and foil produces edge waves, which then become coating cracks at the slitter.
Roll temperature matters as much as line load: 85 °C, 120 °C and 145 °C are all used in production, with hotter rolls improving compaction of PVDF-bonded cathodes by softening the binder, at the cost of a greater tendency to pick and stick.
6. Balancing: why the anode is deliberately 10–25% too big
A full cell is assembled in a fully discharged state, so the cathode must supply every lithium ion that the anode will store — plus the lithium consumed irreversibly in the first cycle to build the SEI. The design parameter is the negative-to-positive capacity ratio:
Commercial cells run 1.1–1.2 for graphite against layered oxides, 1.20 when the cathode has high first-cycle efficiency (LFP, spinel), 1.25 for NCA and older NMC grades with ~88% first-cycle efficiency, and 1.1 for LTO anodes which form no SEI. The reason is lithium plating: if the anode fills before the cathode is empty, the excess lithium deposits as metal on the graphite surface, and that plating is both a capacity loss and a dendrite risk.
Worked example. Take a 3.08 mAh/cm² NMC622 cathode and graphite at 350 mAh/g practical:
So the anode is coated at roughly 104 g/m² against the cathode's 180 g/m². The 12% excess anode capacity is dead mass and dead volume — it contributes nothing to reversible capacity — but it is the cheapest insurance in the bill of materials.
The tolerance interaction is what makes this an operations problem rather than a design problem. With ±0.4 mg/cm² coating tolerance on both electrodes, a nominal N/P of 1.10 can drift to a range of 0.98 to 1.23. A cell at the bottom of that range is plating lithium on a fast charge; a cell at the top is carrying 12% more anode mass than necessary. Both pass a capacity test. Only one passes a 500-cycle fast-charge test.
7. Slitting, assembly and the dry room
Slitting cuts the coated master roll into electrode widths. The failure mode is not dimensional — it is metallurgical. A burr on the cut edge of a 9 µm copper foil can pierce a 12–16 µm polyethylene separator during winding or under thermal expansion in service, creating an internal short that may not show up until the cell has been in a pack for six months.
Production slitting therefore specifies burr height rather than cut width: typical acceptance is under 10–15 µm burr with no fold-over, achieved by matched knife pairs held to ±1 µm thickness tolerance and ≤10 µm total indicated runout, with side pressure and web tension controlled to keep the cut in the ductile regime. Laser cutting avoids the mechanical burr but introduces a 20–50 µm heat-affected zone with localised binder degradation and metal spatter, so it is used mainly for tab cutting and shaped electrodes.
Assembly is either winding (cylindrical, some prismatic) or stacking (pouch, prismatic). Winding is faster and cheaper; stacking gives better volume utilisation and lower tab resistance but requires precise layer alignment — anode overhang beyond the cathode of 1–3 mm on every edge is a hard requirement, because any cathode edge not covered by anode becomes a lithium-plating site.
The entire assembly step runs in a dry room. Cell assembly is specified at dew points of −35 to −45 °C, and electrolyte filling at −60 °C or lower. The numbers behind that: saturated vapour pressure is about 12.8 Pa at −40 °C against 2,339 Pa at +20 °C, so a −40 °C dew point in a 20 °C room is 0.55% relative humidity, carrying about 0.08 g of water per kg of dry air. At −60 °C the moisture content falls to roughly 0.007 g/kg. For reference, ambient air in coastal India holds 15–22 g/kg in the monsoon. The dehumidification plant is therefore removing three orders of magnitude of water, continuously, from a building with people and material moving through it.
8. Electrolyte filling and wetting: the slowest capillary in the plant
Electrolyte is dispensed under vacuum through a dosing lance — typically 1.0–1.5 g/Ah for high-energy cells — with fill accuracy of ±1% by weight, because electrolyte mass is the cheapest way to detect a partially filled cell.
Filling is fast. Wetting is not. Electrolyte must penetrate a porous electrode whose pores are mostly below 100 nm, against a compressed, tortuous, partially hydrophobic structure. The capillary driving pressure is:
For a typical carbonate electrolyte with \gamma \approx 35 mN/m, a contact angle near 25° and a 50 nm pore radius, that is about 1.27 MPa of capillary pressure — plenty of driving force, but the flow rate scales with r^2 through the Hagen–Poiseuille relation, so halving pore radius cuts filling speed by roughly four times. This is why heavy loadings and high compaction are specified with explicit wetting times rather than assumed ones: 12–24 hours under vacuum is normal for large-format cells, and the smallest pores may never fully wet.
9. Formation: three weeks that decide the yield
Formation is the first controlled charge, and its job is to build the SEI — the passivating layer on the anode that stops the electrolyte from decomposing continuously. It consumes 5–12% of the cathode's lithium inventory permanently. That loss is not a defect; it is the cost of a stable interface. A cell with no SEI has excellent first-cycle capacity and a service life measured in weeks.
An industrial formation sequence for large-format cells typically runs:
Stage · Duration · Condition
First wetting · 6–24 h · 25 °C rest
First formation · 10–40 h · 1–2 cycles at 0.1–0.2C
Second wetting · 12–24 h · 25 °C rest
Second formation · 0.5–20 h · 1–2 cycles at 0.2–0.5C
Third wetting · 12–24 h · 40–60 °C
That is 3–7 days for the SEI alone, and the exact sequence — plus the additive package that makes it work — is one of the most closely guarded secrets in the industry. Vinylene carbonate, fluoroethylene carbonate and various sulfones shift the SEI composition and can cut formation time substantially; a competitor can copy a cathode formulation from a datasheet, but not an additive package from an SEM image.
After formation comes degassing (pouch cells swell with ethylene and other by-products and must be vented and resealed) and then aging: one to three weeks at controlled temperature and state of charge, with periodic open-circuit voltage measurement. The measured quantity is self-discharge, expressed as a K-value:
Leakage current immediately after formation runs 20–50 µA/cm², falls to 2–5 µA/cm² within days and below 1 µA/cm² after a couple of weeks. Cells whose OCV drifts faster than the population are the ones with micro-shorts from metal particles, burrs or misalignment — the exact defects that electrical testing at end-of-line cannot see. Sorting on K-value is how a plant removes them before they become a pack-level thermal event.
The capital intensity of this stage is underappreciated. Aging occupies one to three weeks of production in trays, at controlled temperature, with cyclers and chambers that cost as much per square metre as the coating line — and the cells sitting in them are fully converted material that has not yet been sold.
10. The mass audit: where 270 Wh/kg actually comes from
Cell-level specific energy is set by the ratio of active material to everything else. A worked mass model for a 20 Ah NMC811 pouch cell at 4.2 mAh/cm² (cathode 205 mAh/g, anode 350 mAh/g at N/P 1.10):
Component · Mass (g) · Share
Cathode active material · 97.6 · 35.6%
Cathode carbon + binder · 4.1 · 1.5%
Aluminium foil (13 µm) · 8.4 · 3.1%
Anode active material · 62.9 · 23.0%
Anode carbon + binder · 3.3 · 1.2%
Copper foil (8 µm) · 17.1 · 6.2%
Separator (12 µm) · 5.4 · 2.0%
Electrolyte · 55.0 · 20.1%
Tabs, pouch, tape · 20.0 · 7.3%
Total · 273.8 · 100%
At 3.70 V average discharge, 74 Wh over 273.8 g gives 270 Wh/kg. Every gram that is not active material is a direct tax: the copper foil alone is 6.2% of mass and contributes nothing to capacity, which is the entire reason 6 µm foil and tabless cylindrical designs exist. Electrolyte at 20% is the other big non-active mass, and it is set by pore volume — which loops straight back to the porosity decision in section 5.
11. Yield: the multiplier that eats the margin
Cumulative yield is the product of step yields, not the average:
Twelve process steps at 99.5% each give 94.2%. At 99.0% each they give 88.6%. At 98% each they give 78.5%. A plant that looks "99% capable" at every station is losing a fifth of its output.
That is why ramp curves are so painful and so public. Tesla's 4680 line ran at roughly 20% yield in late 2022, at 85 cells per minute against a 350/min target, with contamination sensitivity so high that 0.001% dust caused shorts. Three years later it exceeded one million cells per week. The yield gap between a mature line and a ramping one exceeds any plausible difference in procurement terms — which is exactly the point Argonne's cost modelling makes: at a mature, well-utilised plant, materials dominate; during a ramp, depreciation and scrap dominate.
The cost cascade is easy to compute. A 74 Wh cell at \74/kWh carries \5.48 of cell price. On a 10 GWh line (135 million such cells), 5% scrap is 6.8 million cells — about \37 million a year of cell value destroyed after the value has been added, of which the material content alone is roughly \29 million at 78% materials share.
The yield arithmetic gets ugly because scrap scales as 1/Y - 1. At 95% yield, a plant inputs 10.5 GWh of material to ship 10 GWh and scraps 0.53 GWh. At 20% yield, it must input 50 GWh to ship the same 10 GWh and scraps 40 GWh — 75 times more material destroyed for identical output. At \58/kWh of material and conversion cost, that is roughly \2.3 billion of value incinerated to deliver a 10 GWh year.
12. The dry electrode exception
Everything above assumes wet processing: solvent, drying, recovery, calendering. The alternative is to skip the solvent entirely — mix active material with a fibrillating binder (PTFE), extrude or roll it into a free-standing film, and laminate it to the foil.
Tesla confirmed on its Q4 2025 earnings call (28 January 2026) that both the anode and the cathode of its 4680 cell are made by the dry process at Giga Texas, at more than one million cells per week and roughly 40 GWh of installed capacity. The claimed full-scale economics against wet processing:
- Electrode production cost roughly 50% lower
- Line capital expenditure roughly 41% lower
- Factory energy use more than 30% lower
- Floor space roughly halved
- Line speed 7–10× faster
- Active material fraction 97–98% (versus ~94–96% with binder and conductive additive in a wet formulation)
The remaining constraints are mechanical rather than chemical. Dry films require calender precision in the single-micron range — Chinese equipment suppliers are now quoting ±0.002 mm with >98% yield — and the cathode was the hard part: nickel-rich cathode particles crack and delaminate under the compression needed to fibrillate PTFE, which is why the anode went dry first (Cybertruck cells from late 2023) and the cathode took another two years.
For everyone else, wet processing remains the default: CATL, LG Energy Solution, Samsung SDI and Panasonic all run wet coating in volume, with LG targeting dry lines around 2028. The 4680's energy density still trails the best competitor cells by 10–15%, which is a useful reminder that process innovation and product performance are separate scorecards.
13. What this means in India
India's cell manufacturing build-out is now concrete enough to specify against. The national ACC programme carries a ₹18,100 crore outlay for 50 GWh, and the most recent 10 GWh grid-storage tranche drew 22 bids including Reliance, Amara Raja, Exide, JSW Neo and Waaree.
On the ground:
- Exide has invested about \$632 million and started a 6 GWh lithium-ion cell plant in Bengaluru in September 2026, beginning with e-rickshaw and telecom cells.
- Amara Raja commissioned a ₹500 crore, 60 MWh Customer Qualification Plant at its Telangana Giga Corridor in July 2026, producing cylindrical and prismatic cells across multiple chemistries for OEM validation, ahead of a 2 GWh first line targeted for 2027 within a ₹9,500 crore, 16 GWh programme.
- Agratas is building in Gujarat for Tata Motors and export, and Ola Electric is ramping cell production in Tamil Nadu.
The engineering implication for anyone selling into these plants: the qualification bar is set by the parameters in this article, not by a datasheet's headline Wh/kg. A supplier of electrode tooling, dry-room systems, calender rolls, slitting knives, formation cyclers or cell trays is being measured against ±1.5% coating weight, ±0.05 g/cm³ density, −45 to −55 °C dew point and Cpk of 1.33 or better on each. A CQP exists precisely because those numbers take a year of customer validation to prove.
14. The specification sheet an engineer should actually ask for
If you are buying cells, qualifying a supplier, or designing a pack around a cell, these are the parameters that predict field behaviour — and most of them are absent from a marketing datasheet:
- Areal loading and its tolerance (mg/cm², ±%) — sets capacity consistency and, through porosity, the rate spread.
- Electrode porosity and density (%, g/cm³) — predicts fast-charge behaviour better than any C-rate claim.
- N/P ratio — a value below 1.10 with a graphite anode is a lithium-plating risk under fast charge, regardless of cycle-life claims.
- First-cycle efficiency (%) — determines usable capacity and the anode oversizing required.
- K-value distribution (mV/day) — the real screen for latent internal shorts.
- Separator thickness, porosity and shutdown temperature — the last line of defence against a burr.
- Electrolyte mass per Ah (g/Ah) — an underfilled cell looks fine at beginning of life and fades early.
None of these require a factory visit. All of them require that the supplier has the process capability to hold them, which is the difference between a cell line and a chemistry demo.
The through-line is that a cell factory is a tolerance stack-up exercise with an electrochemical payload. A 0.4 mg/cm² coating variation becomes a 19-point porosity spread, which becomes a rate-capability tail, which becomes a warranty claim three years later in a pack that was thermally designed around a nominal cell. The plants that win are not the ones with the best chemistry. They are the ones whose calender holds ±2 µm at 80 m/min and whose K-value sort removes the 0.3% of cells that would otherwise define their reputation.