Solid-State Batteries

SSB production introduces defect modes with no Li-ion analogue: sulfide-electrolyte moisture reactions that generate H₂S at parts-per-million moisture levels, oxide-electrolyte sintering camber that warps sheets 200 μm across a 200 mm wafer, and composite-cathode compression cracking under 300-700 MPa. Each defect class requires a different detection modality and a different process control loop.

Two Routes, Two Defect Regimes

The oxide route uses garnet-phase LLZO as the solid electrolyte. Tape casting targets film thickness tolerance of ±2 μm. Co-sintering at 1,050-1,200 °C densifies the electrolyte layer while simultaneously bonding cathode material. Lithium-metal anode lamination occurs post-sintering. The cost path to below $150/kWh is feasible at GWh scale according to Schnell et al., but requires sintering yields that pilot data suggests are well below the 90%-plus assumed in published cost models.

The sulfide route uses argyrodite (Li₆PS₅Cl) or similar sulfide electrolytes. Slurry mixing in non-polar solvents avoids moisture exposure during mixing. Slot-die coating followed by calendering at 300-700 MPa compresses the electrolyte layer to greater than 97% relative density, pressures an order of magnitude higher than Li-ion electrode calendering. The minimum practical sulfide electrolyte film thickness is approximately 80 μm; thinner films fracture during handling before cell assembly.

SSB Defect Classes: No Li-Ion Analogue

Oxide route defects include mud-cracking from rapid green-film drying (the same mechanism as Li-ion electrode mud cracking but in a ceramic film that cannot self-heal), camber and warpage during sintering (200 μm warp across 200 mm is the experimentally observed threshold for assembly failure), interfacial impurity phases at the cathode-electrolyte interface from interdiffusion during co-sintering, and residual porosity above 3% that creates ionic conductivity pathways short of the theoretical maximum.

Sulfide route defects include cathode-particle cracking under 700 MPa pressure, the same calendering force required for high electrolyte density fractures the cathode particles below it at pouch scale, a scale-dependent interaction not observed in small-format lab cells. Moisture excursions above 1 ppm H₂O react with sulfide electrolyte to generate H₂S, degrading ionic conductivity and creating safety concerns. Electrolyte film brittleness below 80 μm causes fracture during winding. Lithium-metal anode pitting introduces local current-density hotspots that propagate to dendrite formation.

Yield Reality vs. Cost Models

Published SSB cost models assume sintering yields above 90% and attribute the primary cost driver to electrolyte material cost. Pilot-line experience tells a different story. Sintering yield at early-stage oxide programs is well below 50%, driven by camber and interfacial phase formation that are sensitive to the thermal ramp rate, a parameter not optimized in lab synthesis conditions. Sulfide lines routinely see dry-room moisture excursions as the dominant scrap driver, not material cost.

The gap between published cost model assumptions and pilot-line reality is the yield learning curve, and it is closed by systematic process-defect-outcome data capture, not by materials optimization. The literature optimises materials. The pilot line needs to optimise tolerances and control limits, which requires a different kind of data.

Where SSB Programs Get This Wrong

The scale-dependent defect problem is the most common blind spot. A composite cathode that calendars cleanly in a 25 mm diameter lab cell cracks under 300 MPa at 200 mm × 200 mm pouch scale because the pressure distribution is non-uniform at larger footprints. An oxide electrolyte sheet that shows 10 μm camber in a 50 mm coupon shows 200 μm camber in a 200 mm wafer. These scale-dependent behaviors are not captured in materials journals, they are discovered on the pilot line, and if there is no systematic record, they are rediscovered on every subsequent lot.

The second failure mode is the fragmented data architecture. On most pilot lines, thermal profiles are in the furnace controller log, mechanical data is in the press log, electrochemical data is in the formation cycler, and dry-room moisture is in the building management system. There is no cell-level genealogy connecting them. RCA across these streams takes 2-8 weeks per incident, long enough that the pilot line has run dozens of additional lots before the finding from the first incident is actionable.

What AI Cataloguing Changes for SSB

The highest-value capability for SSB pilot lines is not defect detection accuracy, it is systematic cataloguing of new defect modes as they appear. When a new crack morphology shows up in the sulfide composite cathode, an AI system that logs it with the associated calendering pressure map, moisture level, and eventual cell outcome creates an institutional record that survives engineer turnover and accelerates RCA on every subsequent occurrence. The defect library that builds over a pilot program is worth more than the detection system alone.

Across the thermal, mechanical, and electrochemical data streams, RCA time drops from 2-8 weeks to minutes when those streams are unified under a cell-level ID. The specific question , was this sintering yield event caused by a thermal ramp rate excursion or a tape-cast thickness deviation?, becomes answerable in seconds rather than requiring a manual cross-database investigation.

References

1. 1. Schnell, J., et al. (2019). All-solid-state lithium-ion and lithium metal batteries , paving the way to large-scale production. Energy & Environmental Science, 12(6), 1818-1833. https://doi.org/10.1039/C8EE02692K
2. 2. Schnell, J., et al. (2020). Prospects of production technologies and manufacturing costs of oxide-based all-solid-state lithium batteries. Energy Technology, 8(3), 1901237. https://doi.org/10.1002/ente.201901237
3. 3. Wang, X., et al. (2013). Densification of garnet-Li₇La₃Zr₂O₁₂ solid electrolyte via hot pressing. Journal of the European Ceramic Society, 33(13-14), 2539-2547. https://doi.org/10.1016/j.jeurceramsoc.2013.04.018
4. 4. Singer, D., et al. (2024). Toward manufacturable all-solid-state batteries, defects and their role in yield. Batteries & Supercaps, 7(8), e202400142. https://doi.org/10.1002/batt.202400142
5. 5. White, R.T., et al. (2024). Bridging the gap: in-line quality control for battery manufacturing. Frontiers in Manufacturing Technology. https://doi.org/10.3389/fmtec.2024.1392038