Semi-Solid State Battery Helium Leak and Insulation Screening
I have rejected more semi-solid state battery lots over seal leaks and insulation failures than over capacity shortfalls. Capacity fades slowly and predictably; a 20-micron pinhole in a heat seal kills a pack in weeks. A semi-solid state battery helium leak screening program, paired with disciplined insulation testing, catches those escapes before they reach your customer. In this article I walk through the leak physics, the method trade-offs, the insulation screens, and the acceptance limits I use on production lines.
Why Semi-Solid Cells Fail Seals Differently
A conventional liquid lithium-ion pouch cell carries free electrolyte everywhere. A semi-solid cell holds most of its electrolyte in a gel phase, which changes the leak picture in three ways. First, there is less liquid to weep, so a marginal seal may look clean during a simple pressure-decay check while still admitting moisture and air over months. Second, the gel layer near the seal can bridge a microscopic channel and temporarily mask it, which is why water-vapor ingress, not electrolyte loss, is usually the dominant field-failure mode. Third, semi-solid cells are commonly built with slightly wider heat seals and lower seal temperatures to protect the gel, so the seal itself has a broader process window but more sensitivity to film contamination and wrinkle defects.
The consequence is simple: you cannot rely on visual seal inspection or gross pressure tests. You need a tracer-gas method that measures what actually crosses the seal, and you need to run it on every cell, not on an audit sample. When customers ask me for a custom battery solution with a ten-year service expectation, the end-of-line leak screen is the first control I put on the quotation.
Leak Physics and Units: Getting the Numbers Straight
Leak rates confuse people because the units mix pressure, volume, and time. A helium leak detector reports throughput, typically in Pa-cubic centimeters per second or mbar-liters per second; the two differ by only a factor of 1.333, so for screening purposes they are interchangeable. Helium is the standard tracer because it is inert, its atmospheric background is only about 5.2 ppm, and the smallest atom penetrates micro-channels that air and water barely enter.
Two classes of test exist. A fine-leak test uses a mass spectrometer and detects molecular flow through microscopic paths, with practical detection down to the 1e-7 Pa-cubic cm per second range. A gross-leak test catches paths too large for the fine test: a large hole empties its trapped helium quickly, so the spectrometer sees nothing, exactly like a burst pipe passing a drip check. That is why any real screen runs both, in that order: gross first, fine second. For an aluminum-laminate pouch cell I set an internal fine-leak acceptance of 1e-5 Pa-cubic cm per second or better, and for hermetic prismatic steel or aluminum cans, 1e-6. Those numbers are tighter than most customer datasheets ask for, and the margin is deliberate.
Method Selection: Sniffer, Bombing, and Vacuum Chamber
Three helium methods cover production needs, and choosing wrong wastes both capital and escape rate. The sniffer method pressurizes the cell with helium, then scans the seal with a probe connected to the spectrometer. It is cheap, fast, and localization-friendly, which makes it ideal for pilot lines and root-cause work, but sensitivity depends on scan speed and probe distance, so operator technique dominates. I cap sniffer-based acceptance around 1e-5 to 1e-4 Pa-cubic cm per second in practice.
The bombing method submerges cells in a helium pressure vessel, typically 3 to 5 bar for 1 to 3 hours for laminate cells, forcing helium into any connected channel, then reads each cell in a vacuum chamber connected to the spectrometer. It is quantitative, operator-independent, and the workhorse of end-of-line screening. The vacuum-chamber method without bombing only works for packages backfilled with helium or pre-charged in a helium glove box; it is the fastest option per cell and common on canned cells. A useful hybrid for semi-solid lines: assemble and seal under a dry-air or helium-blended atmosphere, then run vacuum-chamber screening directly and reserve bombing for first-article and lot-qualification checks.
Two cautions from experience. Bombing pressure and time must be qualified per package type; over-bombing a thin laminate can mechanically stress wide heat seals. And watch the fixturing: a chamber that holds forty cells will let one bad cell contaminate the reading of its neighbors through shared vacuum volume, so either fixture cells in isolated pockets or accept slower, single-cell cycling for release decisions.
Insulation and Dielectric Screening at Cell and Module Level
Leak screening catches ingress paths; insulation screening catches internal conductive defects and assembly errors. At cell level I run a dielectric withstand test between shorted terminals and the case or laminate tab, typically 500 volts DC for one second with a leakage limit around 1 milliamp, plus an insulation resistance read at 500 volts DC where I expect 100 megohms or higher on a dry, room-temperature cell. At module level the limits scale with system voltage. Design validation under IEC 62619 and UL 1973 applies insulation coordination and dielectric strength requirements, while ISO 6469-1 gives the familiar vehicle-level minimums of 100 ohms per volt for DC circuits and 500 ohms per volt for AC circuits; I treat those as floors, not targets, and my module release gates ask for 500 volt DC megger readings of at least 100 megohms.
Humidity is the biggest confounder. A cell fresh from a dry room reads beautifully; the same cell after an hour on a humid line can read one order of magnitude lower and still be perfectly healthy. So control the test environment, condition cells before measurement, and track the distribution rather than the mean. A right-tail drift in insulation resistance is the earliest warning of seal problems that the leak detector has not yet caught, which is why I plot both screens on the same weekly chart.
Building the End-of-Line Screen That Catches Escapes
Sequence matters more than equipment brand. My standard order after formation and aging: open-circuit voltage and self-discharge screening first, gross leak, fine leak, then insulation and dielectric tests, then final visual seal inspection under magnification. Running insulation before leak screening wastes time, because a cell that fails gross leak would be scrapped anyway.
Calibrate with a certified helium reference leak, not with the instrument internal check alone; I verify daily at a leak rate within one decade of my acceptance limit. Chamber background matters: after a grossly leaking cell, the vacuum system can stay helium-loaded for tens of minutes, so quarantine cells that fail gross leak and let the chamber purge before trusting the next readings. Record the full leak-rate value, not just pass and fail, per cell, with serial numbers linked. The distribution is your process capability statement; a pass-fail-only record tells you nothing when your escape rate doubles in week nine. Finally, retest rules must be written before the first failure happens: a retested cell gets a fresh bombing cycle, never a second sniff of the same charge, and a failed retest scraps the cell with no engineering waiver on safety-related limits.
What the Standards Actually Gate
Be precise about what each standard covers, because audit findings here are embarrassing. UN 38.3 governs transport and contains no leak-rate test; it assumes your production controls exist. IEC 62619 and UL 1973 cover design validation for industrial and stationary applications, including dielectric strength and insulation coordination, but they validate design, not every unit. IEC 62133-2 applies to portable sealed cells only; do not cite it for rack modules. That means your end-of-line semi-solid state battery helium leak screening limits are internal specifications, and their credibility rests entirely on your documented correlation between leak rate and field reliability. I keep a standing requirement that any acceptance-limit change triggers a review of the last six months of distribution data and, when the line serves aviation or medical customers, a customer notification before the change takes effect.
Frequently Asked Questions
What leak rate should I specify for a semi-solid state pouch cell?
For aluminum-laminate cells I specify 1e-5 Pa-cubic centimeters per second or better by bombing plus vacuum-chamber readout, and 1e-6 for hermetic cans. Treat datasheet numbers above 1e-4 as marketing; require the underlying distribution data instead of a single limit.
Can I use the same helium leak screening equipment for prismatic and pouch cells?
The spectrometer and vacuum chamber are shared, but fixturing, bombing parameters, and acceptance limits are not. Pouch seals deform under clamping, so each format needs its own qualified fixture and a fresh gauge study before its results count.
Why does my insulation resistance reading drift with humidity?
Surface leakage across the seal and terminals scales with the moisture film on the laminate. Condition cells for at least one hour in the test room, keep relative humidity below 60 percent there, and interpret trends only within a controlled environment.
Does helium bombing damage the gel electrolyte?
Not at qualified parameters. Gel electrolyte is essentially incompressible, so 3 to 5 bar for a few hours is benign. The risk is mechanical: thin laminates with wide seals need their own pressure and time study before production bombing is approved.
How often should I calibrate the leak detector?
Daily verification with a certified reference leak near the acceptance limit, full factory calibration every twelve months, and recalibration after any spectrometer filament change. Log verification results so auditors can see drift, not just compliance.
What is the difference between gross leak and fine leak testing?
Fine leak testing finds microscopic paths via trapped helium; gross leak testing finds channels too large to hold a readable helium charge, typically checked by bubbling or weight-change methods. A complete screen always runs both, because neither alone covers the full defect size spectrum.
Further Reading
References
