Hermetic Seal Welding: One Pore Scraps the Part

A weld that looks perfect and a weld that passes a helium leak test are not the same claim. Hermeticity is binary. A sealed enclosure either holds its rated leak rate or it doesn’t, and there is no partially sealed pressure sensor, no mostly-hermetic pacemaker can. One pore that breaks through the joint scraps the part, and it scraps it after the labor, the housing, and everything already inside it are sunk cost. The people searching for this already know what a failed part costs, and they want a process that doesn’t produce one.

This page covers why welds leak, what a vacuum chamber changes about that mechanism, and where the requirement shows up hardest today: sensor housings first, implantable medical devices second. The chamber itself, its size, cost tier, and cycle time, is covered on the qualification page; this page picks up from there and stays on the leak spec.

What Counts as a Hermetic Seal

A hermetic seal is a joint with no continuous path across it for gas or moisture, confirmed by a leak test with a stated pass/fail limit rather than assumed from a clean-looking bead. In practice that means a helium fine-leak test under a named standard, most commonly MIL-STD-883 Method 1014 for electronic packages or MIL-STD-750 Method 1071 for discrete semiconductor devices. Both set acceptance limits that scale with the sealed cavity’s internal volume, because a larger enclosure tolerates a higher absolute leak rate for the same service life. A part that passes visual inspection and fails the leak test has failed, full stop. The inspection that matters is the one with a number attached to it.

Why Welds Leak

A pore is a discontinuity in the metal, and a pore that breaks through from one face of the joint to the other is a physical channel. At that scale, a leak test doesn’t need much of an opening to find it. Two mechanisms produce that kind of porosity in a keyhole weld.

The first is keyhole collapse. During deep-penetration laser welding, the vapor channel the process depends on oscillates, and at atmospheric pressure it periodically collapses. Each collapse traps a bubble of gas that has nowhere to go before the melt pool solidifies around it. The second is dissolved gas: nitrogen, oxygen, and hydrogen picked up from the atmosphere or from surface contamination dissolve into the molten metal, then come out of solution and nucleate pores as the weld cools. Either mechanism, alone or together, produces a joint with pores in it, and some fraction of those pores form a leak path straight through.

A related but separate failure mode is spatter landing on the sealing face itself, on the flange or land the two halves of the housing seat against before the weld even runs. A particle sitting in the joint interface at the moment of fusion becomes a fresh discontinuity, on top of anything porosity contributes. Spatter suppression under vacuum is a real, separately documented effect, and it compounds the porosity story rather than replacing it: a joint face has to be clean of both defect types to hold a leak spec, not just one.

Neither older process solves this cleanly. Resistance seam welding, the traditional route for Kovar and glass-to-metal packages, controls porosity reasonably well but is limited by heat-affected zone width and joint geometry: it works on simple, accessible seams and struggles on complex housing shapes. Atmospheric laser welding is faster and gives finer beam control, but it inherits the same keyhole-collapse porosity risk this section describes, and it still needs a shielding gas that has to physically reach the joint, which is a delivery problem on anything but the simplest geometry. Neither gap is a reason to rule either process out on every part. They’re the reason the porosity mechanism above is worth understanding before picking one.

How a Stable Keyhole Closes the Path

Under vacuum, the keyhole stops oscillating and collapsing the way it does at atmospheric pressure, a mechanism confirmed by direct high-speed camera observation (Reisgen et al., RWTH Aachen ISF, in Lasers in Manufacturing conference proceedings). That removes the first porosity source outright. The second source shrinks too: lower ambient pressure sharply increases the thermodynamic driving force for dissolved gas to escape the melt pool before it freezes, so more of it degasses out instead of nucleating into a pore. Together, keyhole stability and easier degassing are why porosity in steel, titanium, and nickel alloys welded under vacuum drops to very low levels, a mechanism-level finding rather than one machine builder’s number. Fewer pores means fewer chances any one of them lines up into a leak path across the joint.

There’s a second effect worth noting, separate from leak-tightness itself. A part welded shut inside a vacuum chamber traps whatever pressure the chamber held at the moment the seam closes, not ambient air. For a pressure sensor built around a vacuum reference cavity, that’s part of the design intent, not a side benefit. For an implant enclosure, it means no trapped atmospheric gas to outgas against sensitive internals over years in service. It’s a mechanical consequence of the process, worth knowing, not the headline claim.

This page assumes the part already fits inside a production chamber and the buyer already has capital authority for the equipment. The qualification page linked above answers those two questions with the actual chamber envelope and cost drivers — check it first if either one is still open.

The Leak-Rate Number, and Its Limits

One vacuum-welding vendor reports leak rates of 1.2×10⁻⁷ mbar·l/s on hermetic seams in 75 µm stainless foil. That figure is vendor-reported process data, not independently published or peer-reviewed, so treat it as a plausible order of magnitude rather than a spec any buyer can hold a machine to sight-unseen. It sits in the same range as typical fine-leak acceptance limits for small electronic packages under MIL-STD-883, which is why it’s plausible, but a real leak-rate commitment has to come from a test on the actual part, at the actual cavity volume, not a number carried over from someone else’s application.

No ranking page for this term names the actual test methods that decide pass or fail. That is the gap this page fills: a claim of “hermetic” without a stated method and volume-scaled limit is marketing language, not a spec.

Where This Shows Up First: Sensors

Sensors are the strongest current segment for vacuum laser welding, ahead of every other application this process serves. The reason is structural, not incidental. Pressure sensor housings, MEMS packages, and hermetically sealed electronic enclosures are small parts built in batch volumes, on materials, Kovar, stainless steel, glass-to-metal combinations, that already carry a hermetic-or-scrap requirement from the sensor’s own function. A pressure sensor referencing a sealed cavity has no partial-credit version: a housing that leaks doesn’t read pressure correctly, it drifts, and it fails in the field rather than at final test if the leak is slow enough to pass a quick check and fail a real one. Small parts, high unit counts, and an inline helium leak test that fits naturally into a production QA gate are exactly the conditions where a process built around near-zero porosity earns its keep fastest.

Fuel-cell bipolar plates and vacuum-insulated product seals sit in the same category for the same reason: a stack that leaks hydrogen doesn’t run at spec, and hydrogen is small enough to find a leak path that a heavier gas would miss. These are lower-volume than the core sensor segment today, but they share the identical binary requirement and the identical mechanism fix.

The Second Segment: Implantable Medical Devices

Medical device laser welding on implantable enclosures, pacemaker cans, neurostimulator housings, cochlear implant shells, carries the same binary requirement as sensors, plus a consequence sensors don’t have: a weld defect inside the body is also a corrosion-initiation site. A pore or an incomplete-fusion gap gives crevice corrosion somewhere to start, in an environment that won’t forgive a slow failure the way an industrial cabinet might. Titanium (Grade 2 CP, Grade 5 Ti-6Al-4V, Grade 23 ELI) and MP35N cobalt-chromium-nickel-molybdenum alloy are the common enclosure materials, chosen for biocompatibility as much as for weldability, and both take a vacuum-welded seam cleanly, though titanium can still need supplementary in-chamber gas shielding at rough-vacuum pressures to hold weld-zone oxidation in check (Lee, Cheon, Min, and Kim, Int. J. Adv. Manuf. Technol. 123:1297-1305, 2022, DOI 10.1007/s00170-022-10257-5). The regulatory context sits on top of the engineering requirement, not instead of it: FDA 21 CFR 820 design controls and process validation, and ISO 13485 as the quality management framework a device maker’s process needs to sit inside. MIL-STD-883 and MIL-STD-750, named earlier for the leak-test method itself, are microelectronics and discrete-semiconductor test standards, not a device qualification regime. An implant maker doesn’t qualify its process under either one. It qualifies under FDA 21 CFR 820 and ISO 13485, and borrows the same helium fine-leak method those standards define as the test. That’s context a buyer should expect any welding partner to work within, not a claim about a specific machine or vendor’s certification status.

Medical implants sit second in this piece, not co-equal with sensors, for a reason worth stating plainly: sensor housings are the higher-volume, faster-cycling segment where the process earns its keep across more parts per year. Implant welding carries higher per-part stakes and tighter regulatory overhead, which makes it a real segment and a slower one to qualify into. Both share the same binary pass/fail requirement; they differ in volume and qualification timeline, not in what the weld has to do.

Is This Actually Your Problem

Not every leak-tight requirement is actually a hermetic-grade requirement, and confusing the two is an expensive mistake in either direction. A gasketed enclosure that just needs to keep dust and splash out of a control cabinet doesn’t need a welded, leak-tested seam. A part with a real leak-rate spec, tested to a named standard, on an assembly where a leaking unit is scrap rather than a rework ticket, does. The three questions below sort one from the other before a process gets priced.

Chamber size, cost tier, and cycle time work the same way here as anywhere else in this process family — see the qualification page for the vendor envelope data. This page exists to establish that the leak spec is real before that conversation starts.