The Vacuum Stack
Block 3 — Seals & Materials
The wall is not a wall. Under vacuum, every surface leaks gas, every joint is a risk, and every insulator under voltage hides a weak point. This block is about what the chamber is made of, and why the material is the real engineering.
Source rating, used throughout:
- [A] primary: standard (ISO, AVS), peer-reviewed paper, lab or CERN/USPAS.
- [B] manufacturer: technically sound, commercial interest. Cross-check.
- [C] secondary: trade page or recopy. Flagged. Never the only source.
Pressure in Torr (1 Torr is about 1.33 mbar).
1. Two ways to seal
You seal a vacuum joint one of two ways. Elastomer, or metal.
Elastomer (Viton, O-rings). Cheap. Reusable. Demountable in seconds. Good to high vacuum. But it permeates, it outgasses, and it cannot take a hot bakeout. KF and ISO flanges use it. Fine down to the high-vacuum range, no further. [C] HVACO link, flange comparison (KF/ISO use elastomer O-rings), https://hvacolink.com/vacuum-flange/
Metal (copper gasket, ConFlat). A knife-edge bites into a soft oxygen-free copper gasket. The copper flows, fills every machining mark, and work-hardens into a spring. The seal is all metal. No polymer. It holds from atmosphere to below 1e-13 Torr, and it bakes to 450°C. [B] Kurt J. Lesker, ConFlat technical notes (knife-edge, OFHC, -196 to 450°C, < 1e-13 Torr), https://www.lesker.com/newweb/flanges/flanges_technicalnotes_conflat_1.cfm
The trade-off is simple. Copper gaskets are single-use. You replace the gasket every time you open the joint. That is the price of UHV.
2. The ConFlat flange
The ConFlat was invented by Varian about fifty years ago. It is now an ISO standard, ISO 3669. [A] Allectra, UHV practice notes (CF specified by ISO 3669, all-metal, bakeable to 450°C), https://www.allectra.com/notes-for-high-vacuum-practice/
Key facts to keep:
- Body material: stainless 304, 304L, 316L, or 316LN; surface-hardened aluminum exists. 316L costs more but welds better. [A] Allectra; [B] Lesker.
- Gasket: high-purity OFHC copper, standard. Silver-plated copper for systems that bake hot and often, to stop oxide forming. Viton CF gaskets exist for the low-temperature case when you do not truly need UHV. [B] MKS, CF copper gaskets, https://www.mks.com/f/cf-etched-copper-gaskets
- Assembly: tighten bolts diagonally, not around the circle, for an even seal. [B] MKS, CF flanges, https://www.mks.com/f/cf-rotatable-blank-flanges
3. Materials and outgassing
In UHV, the pump fights the walls. The walls release gas, mostly water at first, then hydrogen from the bulk metal. That release is the outgassing rate, and it is what sets your base pressure.
What matters:
- Water dominates early. It sits on the surface and comes off slowly at room temperature. This slow release is the main obstacle to reaching low pressure. [A] CERN, outgassing of vacuum materials (water as the key obstacle, bakeout accelerates its release), https://cds.cern.ch/record/2723690/files/2006.07124.pdf
- Hydrogen dominates late. Once water is gone, hydrogen diffusing out of the steel bulk sets the floor. Vacuum firing at about 950°C for ~24 h drives hydrogen out of stainless and pushes the rate toward 1e-13 mbar·l/s/cm² and below. [A] CERN, same paper; [A] arXiv, seven-chamber comparison (304L, 316L, 316LN, Ti, Al; 316L-XHV and 316LN-XHV vacuum-fired at ~950°C/24h), https://arxiv.org/pdf/2009.10560
- The diffusion energy is measurable. CERN measured ~0.5 eV for a 316LN chamber and ~0.4 eV for OFS copper, both after a 200°C / 20 h bakeout. [A] CERN, same paper.
Material shortlist for UHV: stainless 304L/316L/316LN, OFHC copper, titanium, surface-hardened aluminum. Avoid mild steel and brass. [A] Allectra.
4. Bakeout, the lever
Bakeout is the single biggest lever on base pressure. Heat the whole system during pumpdown. Water leaves the surfaces faster. The pressure rises while hot, then drops far lower once cooled.
Numbers to keep:
- Effective for metals at 12 h or more above 120°C; lower temperature works if you bake longer. [A] CERN, same paper.
- Typical stainless system on a turbo: 200°C. Aluminum can go lower, because Al conducts heat well and steel does not. [A] Allectra.
- Bakeout cuts the outgassing rate by roughly 100x for metals, and the base pressure drops by about the same factor. [A] Allectra.
5. The triple junction — where the connector failed
This is the heart of the matter, and the physics behind the connector story.
A triple junction is the point where three things meet: metal, dielectric, and vacuum. At that point, under voltage, the electric field is much higher than the average field between the electrodes. Electrons get emitted there, by field emission or the Schottky effect. That emission is the start of surface flashover, the conductive path that destroys the part. [A] Wiley, High Voltage, open access, Li 2020 (field at the electrode triple junction far exceeds the average; primary electrons start the flashover), https://ietresearch.onlinelibrary.wiley.com/doi/full/10.1049/hve.2020.0021
Why the field jumps there: the dielectric polarizes. Its polarization charge piles surface charge on the metal at the junction. The enhancement is set by the ratio of the angles the dielectric and the vacuum subtend at the corner. Geometry, not just voltage. [A] AVS, J. Vac. Sci. Technol., French and Pengvanich (field enhancement set by the angle ratio; triple point as flashover initiator), https://www.researchgate.net/publication/235374062_Dielectric_effect_on_electric_fields_in_the_vicinity_of_the_metal-vacuum-dielectric_junction
The fix is geometry and screening. Three lines of defense, all from primary work:
- Shape the dielectric. Sloping the insulator so the cathode angle is obtuse raises the breakdown voltage. Known since the 1960s; accelerators settled on 45° insulators. [A] USPTO patent 9089039 (history of the sloped-insulator discovery, Smith 1964, Shannon 1965), https://image-ppubs.uspto.gov/dirsearch-public/print/downloadPdf/9089039
- Screen the triple junctions. Standard practice in pulsed power is to grow the insulator area to lower the gradient, lengthen the flashover path, and shield the junction where vacuum, insulator, and conductor meet. [A] Phys. Rev. ST Accel. Beams, pulsed HV vacuum insulation, https://journals.aps.org/prab/abstract/10.1103/PhysRevSTAB.10.060401
- Recess the junction. A vacuum gap that recesses the anode triple junction away from the cathode junction measurably improves HV hold-off in feedthroughs. [A] OSTI 2024, statistical study of HV vacuum surface flashover, https://www.osti.gov/pages/biblio/2567870
The lesson behind the connector: a high-voltage feedthrough does not fail because the insulator is too thin. It fails because nobody asked what the field does at the corner where metal, ceramic, and vacuum meet. Under steady DC, charge has time to settle and the geometry decides. The datasheet rating does not.
6. Field abacus — seals at a glance
| Seal | Range | Bakeout | Reusable | Use |
|---|---|---|---|---|
| Viton O-ring (KF/ISO) | atm to high vacuum | low temp only | yes | quick, demountable, HV |
| Viton CF gasket | down to ~UHV edge | limited | yes | low-temp UHV-ish |
| OFHC copper (CF) | atm to < 1e-13 Torr | to 450°C | no, single use | true UHV/XHV |
| Silver-plated copper (CF) | atm to < 1e-13 Torr | hot, frequent | no | repeated hot bakeout |
Sources: Lesker [B], MKS [B], Allectra [A], HVACO link [C].
Method note
Every figure is rated A/B/C. The triple-junction section rests entirely on primary sources (AVS, Phys. Rev., Wiley open access, a USPTO patent, OSTI), because that is the physics that matters most and it deserves the strongest backing. No internal lab data is used. Manufacturer pages are used for hardware facts and flagged as commercial.
Next block: Tubes & Klystrons. From the dispenser cathode to the megawatt tube.