The Entire Process of Lithium Battery Manufacturing
Cathode slurry goes onto aluminum foil. Anode slurry goes onto copper foil. Both get dried, compressed, slit, and assembled into a cell with a separator between them. Electrolyte goes in. The cell gets charged for the first time. Then it sits in a warehouse for weeks while somebody watches the voltage.
That is the process. Below is what goes wrong in it.
The drying oven after the coating station is where electrodes acquire a defect that cannot be detected at the time it forms, cannot be corrected by any subsequent step, and will not manifest until the cell has been in a customer's vehicle for a year or more. PVDF binder dissolved in NMP solvent migrates toward the electrode surface as the solvent evaporates. The foil interface loses binder. Argonne's CAMP facility has published SEM cross-sections with EDS fluorine mapping that show the surface enrichment plainly. The images get shown at every battery conference. The gradient is there on every production electrode dried at production speed.
The QC adhesion test after drying peels from the surface, which is binder-rich, and passes. The interface where delamination starts after cycling cannot be tested nondestructively. It takes cross-section SEM on a destroyed sample, hours of prep, and the result applies only to that one coupon from that one position on the web. The information is forensic, not predictive.
Slowing the oven fixes it. Nobody slows the oven.
A dual-layer approach where a binder-rich primer is coated first, then the main electrode layer, anchors PVDF at the interface. It requires a second coating head or a second pass. Adoption has been limited to a handful of programs where cycle life specifications were extreme enough to justify the throughput penalty.
Thick coatings are a separate and compounding problem. At twenty-five mg/cm² per cathode side, lithium concentration through the electrode thickness during a 1C charge becomes very nonuniform. The layer near the separator delithiates. The layer against the aluminum foil barely participates. NMC secondary particles crack along grain boundaries under the resulting stress. CAMP facility post-mortems show the fractures. Fresh crystal surfaces grow new CEI. Impedance climbs.
Every product team wants twenty-five. Every electrode engineer would prefer seventeen. The product team sets the spec.
PVDF has to fully dissolve in NMP before anything solid goes into the tank. Two to four hours. If it is not fully dissolved, micro-gels form. They are too small for a Hegman gauge to catch. They do not change the viscosity reading. They survive every process step all the way to aging, where they correlate with elevated failure rates. Proving that correlation requires tracking mixing batch identity through coating, calendering, slitting, assembly, filling, formation, and aging, across a gap of three to four weeks, against a background of all the other variables that also affect aging yield. It can take a factory multiple production campaigns to close the loop.
Carbon black pre-dispersion should be a separate high-shear step. Combining it with the main mix saves a vessel and an hour. The agglomerates that survive are below standard QC detection.
NMC811 reacts with NMP above 45°C. Slowly. Generates trace moisture. Can mobilize iron from the mixer walls into the slurry at levels below ICP-OES detection limits. Keeping the mixer below 43°C during active material addition is cheap insurance. Not everyone bothers.
Slurry has six to eight hours of useful life for NMC811 before the rheology drifts.
Raw materials arrive with a certificate of analysis. D50 is always checked. PSD span frequently is not. BET surface area sometimes is not. Both affect downstream behavior in ways that D50 alone cannot predict. A wide-span cathode lot packs differently under the calender and produces different pore geometry. A high-BET lot absorbs more binder than the recipe assumes. Adding span and BET to incoming QC takes twenty minutes of technician time per lot. Factories that have been burned by unexplained coating variability have added them. Others have not been burned yet.
Water washing of NMC811 to remove surface lithium carbonate and hydroxide is now common. The powder must be vacuum-dried immediately and completely afterward. Incomplete drying after washing is worse than skipping the wash entirely, because the wash introduces water into grain boundary spaces that were previously dry.
Graphite, copper foil, separator, electrolyte solvents, conductive carbon: these are design decisions made before manufacturing begins. They are fixed for a given cell program. Manufacturing does not vary them lot to lot. They matter for cell performance and do not present the kind of ongoing process control challenges that mixing and coating do.
Calendering crushes the electrode from about fifty percent porosity down to the mid-twenties to mid-thirties. Porosity is measured. Tortuosity, which governs ion transport rate and therefore fast-charge capability, is not measured in production. FIB-SEM gives it in the lab. Hours per sample. The two quantities are correlated imperfectly enough that controlling porosity alone may not be sufficient for cells targeting very high charge rates. For moderate charge rates it has been adequate.
Slitting, tab welding, stacking, winding: mechanical operations. Edge burrs from slitting can pierce separators months later. Laser slitting eliminates them. CATL uses laser slitting on premium cells. Stacking alignment for pouch cells needs to hold within half a millimeter. Vision systems check every layer.
Electrolyte goes in under vacuum in a dry room below minus 40°C dew point. Seconds to inject. A day or two for the liquid to permeate the compressed electrode stack by capillary action. Every cell sits in a rack during that soak. The warehouse space is the least discussed and one of the most expensive parts of a battery factory.
The dry room's dehumidification system runs on compressed air supplied by oil-free screw air compressors, typically Class 0 rated under ISO 8573-1 to guarantee zero oil contamination in the air stream. These compressors feed the desiccant rotor dehumidifiers that maintain the minus 40°C dew point. A compressor failure or a pressure drop in the supply line lets ambient moisture flood the dry room within minutes, and any open cells inside are contaminated. Redundant compressor banks with automatic failover are standard, and compressed air dew point is monitored continuously at multiple points downstream of the dryer.
Fill volume is set ten to thirty percent above electrode pore volume. The excess sustains SEI repair over the cell's service life. Dialing it in precisely takes months of cycling data from sample cells. New programs launch with conservative fills that get trimmed over time.
The additive package in the electrolyte (VC, FEC, plus a proprietary combination of sultones, borates, and alternative lithium salts specific to each manufacturer) can double cycle life. Additive formulation teams are small, secretive, and among the highest-paid groups at a cell maker.
Dry room opex is eight to twelve percent of total factory operating cost.
The first charge creates the SEI. The SEI that forms during this charge is the SEI the cell will carry for its entire life. It does not improve afterward. C/20 formation produces denser, more LiF-rich SEI than C/5. The self-discharge gap between C/20-formed and C/5-formed cells is roughly a factor of two and it is permanent. This has been reproduced across electrolyte systems and graphite types and cell formats and is not particularly controversial.
The rest period after the formation charge matters in a way that is easy to undervalue because it looks like idle time. Metastable organic SEI components rearrange toward stable inorganic phases during rest. Twelve to twenty-four hours of rest after formation is standard. Cutting it to four hours saves twenty hours of channel time per cell across fifty to a hundred thousand channels. The throughput incentive is severe. Cells formed with truncated rest show faster impedance growth during their first few hundred cycles in the field. The SEI that was not given time to stabilize continues rearranging during use, opening graphite surface, consuming lithium, generating gas.
Formation equipment for a 10 GWh factory is fifty to a hundred thousand independent channels. Over a hundred million dollars in capital.
Pouch cells get degassed after formation. The pouch is opened under vacuum, gases evacuated, resealed.
Aging is the last gate. Cells sit one to four weeks. OCV monitored. Decay above three to five millivolts per week trips rejection. What aging catches: a fifteen-micron iron particle from a worn slitting blade or a contaminated mixing batch, sitting inside the cell since electrode fabrication, now dissolving electrochemically after the cell's first charge, migrating, depositing across the separator, slowly building a conductive bridge. Two to three weeks for the voltage signal to become detectable.
Mature lines reject one to three percent at aging. New lines reject ten to fifteen percent.
The aging rejection rate is the KPI that factory management watches because it is the integral of every upstream defect source collapsed into one number delivered weeks after the defects were created.
Grading after aging: capacity and resistance measured, cells sorted into matched bins of one to two percent capacity spread for EV packs.
Total process time about forty-five days. Active processing five to seven days. The rest is waiting.