The Big Machines

A laser pulse hits a photocathode inside an RF gun, and the freed electrons are accelerated at once by a strong RF field. The vacuum here is brutal, because the cathode is fragile. The LCLS-II injector (LBNL, out of the decade-long APEX program): a normal-conducting CW gun at 185.7 MHz, a 1.3 GHz buncher, and a gun load-lock that changes photocathodes under vacuum. [A]

Figure	Value
APEX/VHF gun	20 MV/m, ~750 kV, ~3e-10 Torr
Gun requirement	low 1e-9 Torr; O₂ partial < 1e-11 to keep Cs₂Te alive
O₂ 1/e cathode life (APEX)	~14.5 Langmuirs ≈ 402 hours

During commissioning, slow NEG activation cut the hydrocarbon partial pressure, killed multipacting, and unlocked the full 80 kW of RF. [A]

Every decade buys cathode life. Life is measured in Langmuirs (1 L = 1e-6 Torr·s), that is the whole point of UHV here. The recurring failure is the tubes block again at higher voltage: cathodes die by arcing and laser damage, and the mounting controls both vacuum and RF.

The beam enters a 4 GeV linac on 1.3 GHz superconducting RF: ~100 MeV, 100 pC, 12 A bunches, normalized emittance ~0.4 µm, rates to 1 MHz. [A]

Cold is a pump. SRF cavities run at 2.0 K, and cold walls are near-perfect cryopumps. At this scale cryogenics and vacuum stop being two problems and become one. The LCLS-II-HE upgrade pushes it further with an SRF low-emittance gun. [A]

Two proton rings, 7+7 TeV, in the 26.7 km LEP tunnel, 8.3 T magnets at 1.9 K, the beam vacuum inside a magnet cold bore that acts as a cryopump. [A] Two systems, very different:

System	Requirement	Scale
Insulation vacuum	~1e-6 mbar to stop heat conduction	50 km, 15,000 m³; cryopumps to 1e-4 Pa once cold
Beam vacuum	days of beam lifetime → ≥3 decades better	54 km UHV; 48 km cryogenic at 1.9 K, 6 km room-temp NEG

The elegant part is the beam screen: a perforated screen (4% transparency) inside the 1.9 K bore, run at 5–20 K, intercepting synchrotron radiation (~45 eV critical at 7 TeV) and image-current heat so they do not load the 1.9 K system, while the slots still let gas pump through to the cold bore; in-situ heating to 90 K flushes condensed gas. [A]

The scale made concrete: 780 ion pumps, 170 Bayard-Alpert gauges, 1084 Pirani and Penning gauges. NEG coating was born and industrialized at CERN. All beam vacuum leak-tight to better than 1e-11 mbar·l/s helium. [A]

Same rule as a single ConFlat. The largest machine in the world comes down to a leak-tight, clean, contamination-free surface. Scale does not change the physics. It multiplies it by 54 km.

A single ranking of accelerators by pressure is a category error. Vacuum quality is not one number, and the figure of merit changes from one machine class to the next: a storage ring arc, a superconducting cavity string, and an antimatter trap cannot be compared on a pressure axis. The map at the top of this page colors each site by the figure of merit that is actually hard, not by how low the pressure reads.

Class	The hard problem	Pressure regime
Antimatter / XHV	reaching and holding extreme high vacuum	below 1e-12 Torr
SRF linacs	particulate cleanliness inside the cavity	cavity XHV
4th-generation rings	distributed pumping in tiny apertures	about 1e-10 Torr
Hadron colliders, factories	dynamic vacuum at scale	cryogenic, beam-induced
3rd-generation rings	classical UHV with lumped pumps	1e-9 to 1e-10 Torr

Antimatter. The apex, and not close. Antiproton decelerators such as ELENA and the AD complex at CERN hold extreme high vacuum so that antiprotons survive without annihilating on residual gas. The Penning traps go further still, into the most extreme vacuum ever produced on Earth, reached by cold cryopumping inside a sealed can. Nothing else in accelerator physics operates within several decades of it.

SRF linacs. The cavity vacuum is extreme by construction, but pressure is not the metric here. A single micron-scale particle landing on the niobium surface seeds field emission and kills the accelerating gradient, so the contest is fought in cleanrooms, with built-to-print discipline and technicians who move slowly to avoid shedding particles. European XFEL is the reference. LCLS-II and its high-energy upgrade carried that TESLA and XFEL lineage into continuous-wave operation, detailed in the tabs above; SHINE in Shanghai is now building the largest CW superconducting linac in Asia, with FRIB, ESS, and CEBAF rounding out the family.

4th-generation rings. The most interesting class technically. Ultra-low emittance forces small-aperture magnets, so the vacuum chambers shrink until no conventional pump will fit through the bore. The answer is to coat the entire inner wall with non-evaporable getter, turning the wall itself into a distributed pump with no conductance penalty. MAX IV proved it could be done across an entire ring, with activation near 180 °C (356 °F). Sirius is the one to respect most, because the Brazilian laboratory built a sovereign coating capability from near zero. ESRF-EBS was the first major rebuild, and APS-U and HEPS are the large machines now in commissioning, with PETRA IV, ALS-U, HALF, and a full rebuild wave behind them.

Hadron colliders. For the LHC the static number is not the story, as the tab above lays out. The story is dynamic vacuum at scale, and non-evaporable getter coating was invented here, at CERN, for exactly this problem. The high-luminosity upgrade adds amorphous-carbon coatings to suppress the electron cloud on the beam screen.

Particle factories. The luminosity frontier, led by SuperKEKB, runs enormous beam currents, so the vacuum is dominated by beam-induced effects and antechamber engineering rather than base pressure.

3rd-generation rings. The classical ultra-high-vacuum base on which the whole 4th-generation revolution was built: aluminium or stainless chambers, lumped ion and getter pumps, and an antechamber to absorb the photon fan. SPring-8, NSLS-II, Diamond, SOLEIL, ALBA, SSRF, SOLARIS, and the Australian Synchrotron are the working backbone.

This know-how is not spread evenly. A handful of institutions hold it and train everyone else, and that concentration is the strategic point of the whole block.

Institution	Strength
CERN, TE-VSC	the global reference; invented NEG coating; trains the rest of the field
STFC Daresbury, ASTeC	NEG research and the reference literature
DESY	both superconducting RF (XFEL) and rings (PETRA IV)
FNAL, JLab, CEA Saclay, KEK	the cavity and cryomodule consortium
ESRF	NEG on insertion devices, the historical pioneer
Argonne, LBNL, BNL	the US 4th-generation builds
LNLS (Sirius)	a sovereign coating capability built from scratch
IHEP, SARI	China’s rapid ascent, with HEPS and SHINE

The frontier is not one place. The lowest pressure ever made sits in antimatter traps. The hardest distributed pumping is in the diffraction-limited rings. The hardest cleanliness is in the superconducting linacs. The hardest dynamic vacuum at scale is the LHC. Forced into a single composite of engineering depth and expertise, CERN leads clearly, followed by the DESY and diffraction-limited-ring block, then the superconducting RF consortium, with the Chinese programs closing fast.

The strategic point is the last one. This know-how is concentrated in very few places, and it is exactly what private fusion and quantum hardware now need, without forty years of accumulated practice behind them. The scarcity is structural, not temporary.

Machine	Vacuum	Cold?	Pumping	Vacuum is for
RF photoinjector (LCLS-II gun)	~3e-10 Torr, O₂ < 1e-11	no (NC gun)	ion + NEG, load-lock	photocathode life
SRF linac (LCLS-II)	UHV, cavities at 2.0 K	yes	cold walls cryopump	superconductivity + beam
Synchrotron / FEL beamline	UHV, 1e-9 to 1e-10	partly	ion, NEG, sublimation	beam lifetime, clean light
Hadron collider (LHC beam)	UHV over 54 km, screen 5–20 K	48 km at 1.9 K	cold bore + 780 ion pumps + NEG	beam lifetime in days

A vacuum career is one idea at every scale. The mean free path sets the regime. The pump fights the wall. The gauge reads what you cannot see. The seal and the material decide whether you reach base pressure. The joint holds it permanently. The tube turns a beam into power, and dies at the triple junction if you ignore the field at the corner. The machine is all of it, leak-tight and clean, multiplied by kilometres. The physics does not get easier as the machine grows, it gets repeated, more times, with less room for error. That is the job.