Vacuum leaks rarely come from the machined surface
When a chamber refuses to reach base pressure, the sealing surfaces are almost never the problem. The leak sits in a weld, in a fitting joint, or in a volume that never gets pumped out at all. That is why vacuum parts are ordered as welded assemblies far more often than people expect, and why the welding process deserves a line on the drawing instead of a phone call.
The test result is a number, not an opinion. Helium leak testing gives you a rate, and that rate is what separates a part that works from a part that keeps a diffusion pump busy.
What makes a seam vacuum-tight and what does not
Typical acceptance values are a helium leak rate of 1×10⁻⁶ mbar·l/s or better for HV work, and 1×10⁻⁹ mbar·l/s or better for UHV, often 1×10⁻¹⁰ on semiconductor chambers. A weld does not reach those numbers by looking good. It reaches them through process choice and full penetration.
- TIG with full penetration and a back purge for chambers, tubes and flange joints
- Electron beam welding when the joint is deep and narrow and distortion has to stay small
- Laser welding for thin-wall and small geometry work
- Not MIG: spatter inside a vacuum volume becomes a virtual leak and a particle source
- No sealant, no tape and no low-melting filler — anything that outgasses under bakeout is out
Penetration, porosity and virtual leaks
A weld that passes a visual check can still fail a helium test. Porosity is internal, so nobody sees it, and a partial-penetration root leaves a crevice that keeps releasing gas for days. Both are process issues, not machining issues, which is why the weld process belongs in the specification.
The other trap is the virtual leak. A trapped volume — a blind tapped hole, a doubled wall, an unwelded gap between two plates — slowly bleeds gas and behaves exactly like a real leak. The fix is in the design: vent holes, drill through instead of blind, weld both sides wherever a tool can reach.
Weld first, finish the sealing faces afterwards
Welding puts local heat into the assembly and the joint contracts as it cools. A sealing face machined before welding will not be flat after it. The sequence below is the one that holds flatness and surface finish where they matter.
| Step | Why it has to happen in this order |
|---|---|
| Rough machine | Leave 0.3-0.5 mm on sealing faces for the post-weld cut |
| Weld | Full penetration, back purged, both sides where accessible |
| Stress relieve | If the drawing specifies it, do it before the finish cut, not after |
| Finish the sealing faces | Flatness and Ra are established after the weld, never before |
| Clean | Pickle, passivate, ultrasonic clean — weld scale and swarf both outgas |
| Leak test | Helium at the rate on the drawing, recorded per assembly |
Materials, plating and cleanliness
- 304L and 316L for welded vacuum assemblies — the low carbon grade resists sensitization at the weld
- 6061 aluminium welds acceptably but is prone to porosity; 7075 should not be welded into a vacuum
- Copper is standard for CF flange knife edges; keep plated fasteners out of the vacuum volume
- No zinc or cadmium plating inside a chamber, no matter how convenient at assembly
- State the bakeout temperature (150-250 °C is common) — it eliminates most polymers and elastomers
- Fingerprints outgas. Gloves and a final clean before the leak test are part of the process
What to put on the drawing
- Leak rate and test method: helium, vacuum mode or sniffing, with the number you will accept
- Weld process and full penetration, welded both sides where the joint allows it
- Vent holes, and no blind hole opening into the vacuum volume
- Sealing faces: flatness and Ra after welding, with the note weld then finish machine
- Bakeout temperature, and any zones that must stay unplated
- Cleaning and packaging: sealed after cleaning, nitrogen blanket if the part travels far
TruPart Precision is an ISO 9001:2015 shop in Dongguan. We machine vacuum hardware in 304L, 316L and 6061 — chamber bodies, flange fittings, feedthrough housings and base plates — to ±0.005 mm on critical features, with sealing-face flatness and Ra measured on the CMM and reported with the batch.
Frequently Asked Questions
What helium leak rate do I need for UHV?
For UHV work, specify 1×10⁻⁹ mbar·l/s or better, and 1×10⁻¹⁰ on semiconductor chambers. HV assemblies normally accept 1×10⁻⁶ mbar·l/s. Always state the test method as well, because a sniffing test and a vacuum-mode test do not measure the same thing.
Can you weld 7075 aluminium for a vacuum chamber?
It should not be welded for vacuum service. 7075 is prone to hot cracking and porosity, and porosity inside a vacuum weld is a leak you cannot see. Use 6061 for welded aluminium assemblies, or switch to 304L or 316L if the joint has to be reliable.
Why does my chamber take days to pump down?
Most often it is a virtual leak, not a real one: a trapped volume that slowly releases gas. Blind tapped holes, doubled walls and unwelded gaps between plates all behave like a leak but pass a helium sniff. Vent holes and through-drilling usually fix it.
Is TIG welding good enough for a vacuum chamber?
For chambers, tubes and flange joints, yes, as long as the weld is full penetration and back purged. Electron beam welding is the better choice when the joint is deep and narrow or when distortion has to stay minimal.
Do sealing faces really have to be machined after welding?
Yes. Welding distorts the assembly, so a face that was flat before the weld will not be flat after it. The usual specification is to leave 0.3-0.5 mm of stock, weld, then finish machine the sealing faces and verify flatness and Ra before the leak test.