⚗️ Build 09 · Vacuum Systems Guide

Ultra-High Vacuum

Before you trap a single atom, you must first remove 1023 of them. The vacuum system is the silent foundation of every AMO experiment — invisible, yet its failure destroys everything above it in minutes.

10⁻¹⁰
Target pressure (mbar)
2.7×10⁶
Molecules/cm³ at UHV
~500 km
Mean free path at UHV
~1000 s
Atom lifetime at 10⁻¹⁰ mbar

01 Why Vacuum?

The physics of the invisible enemy — and 400 years of fighting it

The fundamental problem: background gas collisions

A laser-cooled atom or molecule confined in an optical tweezer is cold — maybe 5 μK — but the residual gas in your chamber is at room temperature (~300 K). When a hot N₂ molecule (moving at ~470 m/s) collides with your precious cold atom, the energy transfer easily exceeds the trap depth, and the atom is ejected in microseconds. The collision rate per trapped atom is:

$$\Gamma_{\rm bg} = n_{\rm bg}\,\sigma_{\rm eff}\,\bar{v} \qquad \Rightarrow \qquad \tau_{\rm trap} = \frac{1}{\Gamma_{\rm bg}} = \frac{k_BT}{P\,\sigma_{\rm eff}\,\bar{v}}$$

where $n_{\rm bg} = P/k_BT$ is the background number density, $\sigma_{\rm eff} \approx 5 \times 10^{-15}$ cm² is the effective loss cross-section for a hot atom on a cold atom, and $\bar{v} \approx 470$ m/s for N₂. At $P = 10^{-8}$ mbar the trap lifetime is only ~10 s. At $10^{-10}$ mbar it reaches ~1000 s — long enough for a multi-hour experiment.

A very brief history

1643
Torricelli inverts a mercury-filled tube into a dish — the mercury column falls to 76 cm, and the empty space above it is the first reproducible vacuum. "Nature does not abhor a vacuum," he concludes: it weighs about one kilogram per cm².
1654
Otto von Guericke evacuates two bronze hemispheres (30 cm diameter) with his newly invented piston pump. Two teams of eight horses cannot pull them apart — a performance staged for Emperor Ferdinand III. Atmospheric pressure is ~10⁵ N/m². The Magdeburg hemispheres remain the most theatrical experiment in physics.
1855
Geissler invents the mercury displacement pump, reaching ~10⁻¹ mbar — enough for glowing gas discharge tubes and, eventually, neon signs.
1879
Edison's lightbulb requires 10⁻⁴ mbar to keep the carbon filament from oxidising. A functioning vacuum system was the literal prerequisite for electric light.
1895–97
Röntgen discovers X-rays and J. J. Thomson discovers the electron — both inside vacuum tubes. The two defining discoveries of modern physics happened in evacuated glass vessels.
1960s
Turbomolecular and ion pumps developed for particle accelerators (CERN, Brookhaven) push the frontier to 10⁻¹¹ mbar. UHV becomes routine in surface science.
1990s–
AMO physics adopts UHV as standard. BEC (Cornell & Wieman, 1995), optical tweezers, and Rydberg arrays all live at 10⁻¹⁰–10⁻¹¹ mbar.

Trap lifetime vs pressure

The difference between 10⁻⁸ mbar and 10⁻¹⁰ mbar is 100× in pressure — and 100× in atom lifetime. Most experiments need at least several minutes of trap lifetime to run a meaningful protocol, so UHV is non-negotiable.

10⁵s 10⁴s 10³s 10²s 10s -11 -10 -9 -8 -7 log₁₀(P / mbar) ~1000 s ~10 s UHV
💡 Rule of thumb: Every decade of pressure improvement multiplies your atom lifetime by 10. Going from 10⁻⁸ to 10⁻¹⁰ mbar turns a 10-second lifetime into a 1000-second one.

02 The Pressure Landscape

Drag the slider to explore number densities, mean free paths, and what lives at each pressure

Interactive Pressure Explorer

XHV 10⁻¹² UHV 10⁻⁹ HV 10⁻⁵ Atm 10⁵
Ultra-High Vacuum
Pressure
10⁻¹⁰ mbar
10⁻⁸ Pa
7.5×10⁻¹¹ Torr
Number density
2.4×10⁶
molecules / cm³
Mean free path
~500 km
for N₂ at 293 K
Atom lifetime
~1000 s
background-gas limited
Regime Pressure range (mbar) n (cm⁻³) Mean free path Example applications
Atmosphere ~10³ 2.7×10¹⁹ ~68 nm Ambient lab air
Low Vacuum 10³ – 1 10¹⁹ – 10¹⁶ nm – μm Rough pumping, CVD
Medium Vacuum 1 – 10⁻³ 10¹⁶ – 10¹³ μm – mm Freeze-drying, vacuum ovens
High Vacuum 10⁻³ – 10⁻⁶ 10¹³ – 10¹⁰ mm – km Thin-film deposition, SEMs
Ultra-High Vacuum 10⁻⁶ – 10⁻⁹ 10¹⁰ – 10⁷ 10 – 10⁴ km MOTs, surface science, particle accelerators
XHV / AMO < 10⁻⁹ < 10⁷ > 10⁴ km Optical tweezers, BEC, Rydberg arrays
$$\lambda = \frac{k_B T}{\sqrt{2}\,\pi\,d^2\,P}$$

Mean free path for N₂ ($d \approx 3.7$ Å, $T = 293$ K). At $10^{-10}$ mbar: $\lambda \approx 500$ km — longer than the flight from Purdue to New York.

03 The Pumping Chain

No single pump spans 15 decades of pressure — you need a staged cascade

CHAMBER 10⁻¹⁰ mbar ~2.4×10⁶/cm³ ION PUMP 10⁻⁵→10⁻¹¹ mbar no exhaust NEG / TSP H₂ getter TURBO PUMP 10⁻²→10⁻⁸ mbar ROUGH PUMP 10⁵→10⁻² mbar atm GV

Animated gas flow (→) shows the pumping direction. GV = gate valve (isolates the turbo from the chamber during venting or pump failure). Ion pump and NEG/TSP are always-on; turbo + rough are the initial pumpdown path.

🔩
Stage 1 — Rough Pump (Scroll or Diaphragm) Brings the system from atmosphere to backing pressure
10⁵ → 10⁻² mbar

The rough pump is first in line. It operates by mechanical compression — gas is drawn in, physically compressed, and pushed out to atmosphere. Two designs dominate AMO labs:

  • Scroll pump (recommended): Oil-free, two interleaved spiral scrolls orbit to compress gas. No oil mist, no vibration from pistons. Leybold, Agilent, and Edwards make reliable versions used in most tweezer labs.
  • Diaphragm pump: Even simpler, no sliding contacts — a flexing membrane compresses gas. Lowest ultimate pressure (~5 mbar), best for clean low-throughput applications.
  • Rotary vane (legacy): Uses oil-lubricated vanes. High throughput but oil backstreaming can contaminate your chamber. Always use a foreline trap if you have one.
Ultimate: ~5×10⁻³ mbar Speed: 5–30 m³/h Always backed to atmosphere
💡 Always start with the rough pump. Never turn on a turbo against atmosphere — the spinning blades will vaporize on contact with air molecules at their operating speed of 50,000–90,000 RPM.
🌀
Stage 2 — Turbomolecular Pump The workhorse of high and ultra-high vacuum
10⁻² → 10⁻⁸ mbar

A turbo pump is essentially a very precisely machined turbine. Blades spin at 30,000–90,000 RPM — fast enough that the blade speed ($\sim$300 m/s) is comparable to the mean thermal velocity of residual gas molecules. Molecules that hit the blades are imparted a directed momentum toward the backing port; molecules trying to re-enter from the backing side are physically swept away.

The pumping mechanism is purely kinetic (molecular flow regime). This is why a turbo needs the rough pump to first bring the pressure below ~10 mbar — at higher pressures the mean free path is shorter than the blade spacing and viscous flow prevents the turbo from working.

  • Compression ratio for N₂: ~10⁹; for H₂: ~10³ (lighter molecules are harder to pump).
  • Tip seals: Modern turbos use magnetic bearings (vibration-free, no oil). Ideal for optical experiments where vibration couples to trap displacement.
  • Gate valve: Always place a gate valve between turbo and chamber. If power fails, the gate valve closes and protects your UHV from the backstreaming turbo.
Speed: 50–500 L/s RPM: 30k–90k Backed by: rough pump
⚠️ Never vent a spinning turbo to atmosphere. The blades are thin-walled aluminium rotating at near supersonic speeds. A vent valve on the turbo's high-vacuum port controls this — open it slowly through a restricted needle valve after powering down the pump.
Stage 3 — Ion Pump Solid-state, no moving parts, no vibration — the UHV workhorse
10⁻⁵ → 10⁻¹¹ mbar

Ion pumps have no mechanical parts. They work by ionizing residual gas atoms and then burying the ions inside the pump material. A strong magnetic field ($\sim$0.1 T) traps electrons in long helical orbits, massively increasing their path length and ionization cross-section. The resulting ions are accelerated into titanium cathode plates where they are implanted and chemically bound (chemisorption).

  • Pumping mechanism: Penning discharge. Residual gas is ionized; ions slam into Ti cathodes and are embedded. Reactive gases (O₂, N₂, CO) pump well; noble gases (Ar) pump poorly and can cause "argon instability" — avoid noble gas leaks.
  • Current readout: The ion pump current is proportional to pressure — it doubles as a gauge. A 20 L/s pump at 10⁻¹⁰ mbar draws ~1 nA of current.
  • Magnetic field: The external magnet stacks make ion pumps heavy (~5–10 kg for a 20 L/s pump). The stray field ($\sim$1 mT at 10 cm) can affect magnetic-field sensitive experiments — account for this in your layout.
Speed: 2–200 L/s Power: ~5 W at UHV HV supply: 3–7 kV
The gold standard: Once you've baked and reached 10⁻⁸ mbar with the turbo, close the gate valve, start the ion pump, and leave it running indefinitely. Ion pumps get better with time as they outgas themselves.
🧲
Stage 4 — NEG Pump & Titanium Sublimation Chemical getters for H₂ — the molecule ion pumps hate
XHV supplement

Ion pumps have poor compression ratios for H₂ (the dominant residual gas at UHV). NEG (Non-Evaporable Getter) pumps and TSP (Titanium Sublimation Pumps) address this through chemisorption — gas molecules stick permanently to a reactive surface.

NEG cartridges (e.g., SAES St707: Zr-V-Fe alloy) adsorb H₂, CO, N₂, and O₂ at elevated temperatures or passively near room temperature after activation at ~400°C. They have zero magnetic field, zero vibration, and zero power in passive mode — ideal for sensitivity-limited experiments.

TSP: An electrical current heats a Ti filament until Ti evaporates and coats the chamber walls. The fresh Ti film aggressively chemisorbs nearly all reactive gases. A single filament firing lasts ~30 min; TSPs are typically operated every few days to maintain XHV. They can achieve transient pumping speeds of thousands of L/s.

NEG speed: 5–100 L/s for H₂ TSP filament: ~40 A for 45 s No moving parts

04 Baking Out — Cleaning from the Inside

The single most important ritual before first pump-down

Why does every surface hate you?

A freshly machined stainless steel surface has roughly one monolayer of water adsorbed on every exposed face — about $10^{15}$ H₂O molecules/cm². A typical AMO chamber has perhaps 5,000 cm² of internal surface. That's $5 \times 10^{18}$ water molecules slowly evaporating into your vacuum over days and weeks. At room temperature, the thermal desorption rate is so slow that it would take years to pump all of that water out through the turbo. Baking dramatically accelerates desorption: the outgassing rate scales roughly as $e^{-E_a/k_BT}$ where $E_a \approx 0.5$ eV for physisorbed water. A modest 150°C bake speeds outgassing by a factor of ~10⁶.

The bakeout pressure curve

10⁻⁴ 10⁻⁶ 10⁻⁸ 10⁻⁹ 10⁻⁷ Pressure (mbar) 0h 8h 24h 48h 72h cool Bake ON 200°C Bake OFF 10⁻¹⁰ T ≈ 200°C

Bakeout protocol

Typical schedule

1. Preparation: Remove all heat-sensitive parts: viton O-rings (max ~200°C), elastomers, electronics boards, optical coatings on unprotected elements. Loosen flange bolts slightly to allow thermal expansion (re-torque during cool-down).

2. Ramp up: Heat to 150–250°C at <5°C/min. Fast ramps crack CF gaskets. Wrap the chamber in aluminium foil + fiberglass heating tape. Thermocouples on flange bodies, not heating tape.

3. Hold: 24–72 hours at temperature. Pressure will spike, then slowly decrease. The turbo (or a dedicated bakeout turbo) should run the whole time. Do not close the gate valve during bake.

4. Cool down: <3°C/min. The biggest pressure drop occurs as you cool below 100°C — remaining water re-adsorbs on any cold surface outside the chamber, leaving your inner surfaces clean.

⚠️ Viton O-rings degrade above ~200°C. CF flanges (knife-edge + Cu gasket) are UHV-grade to 450°C and are always preferred.
🚫 Never bake with an ion pump running. Ion pumps have internal Alnico magnets that demagnetize above ~100°C. Start the ion pump only after the chamber has cooled and the turbo has pumped to <10⁻⁷ mbar.

05 Materials & Connections

Not everything belongs inside a UHV system — material choice is irreversible

CF vs KF flanges

PropertyCF (ConFlat)KF (ISO-KF)
Seal typeCu knife-edge gasketViton O-ring + centering ring
Pressure range10⁻¹² mbar → 10 bar10⁻⁸ mbar (practical)
Bakeout✓ up to 450°C✗ max ~120°C (viton)
ReuseGasket single-useReusable if clean
Use in AMOMain chamber, sourcesForeline, RGA ports
💡 CF torque sequence: Star pattern, 4–5 passes with increasing torque. Final: typically 14–20 N·m for SS bolts on 2¾″ CF. Under-torquing leaks; over-torquing damages the knife edge.

Outgassing rate comparison

MaterialOutgassing rate*AMO suitability
304/316L SS (electropolished)~10⁻¹² mbar·L/s·cm²Excellent
OFHC copper~10⁻¹² mbar·L/s·cm²Excellent (gaskets)
Fused silica / quartz~10⁻¹² mbar·L/s·cm²Excellent (windows)
Aluminium (bare)~10⁻¹⁰ mbar·L/s·cm²Acceptable; avoid bake
Viton~10⁻⁸ mbar·L/s·cm²KF only; not baked
Kapton (wires)~10⁻⁹ mbar·L/s·cm²Acceptable in low flux
Regular rubber / PVC>10⁻⁶ mbar·L/s·cm²Never use in UHV
Zinc / cadmiumHigh vapor pressureNever use

*After 24h bakeout at 200°C. Values highly dependent on surface preparation.

Copper gaskets — handle with care

CF gaskets are OFHC (oxygen-free, high-conductivity) copper — soft enough for the SS knife edges to bite into. Key rules:

  • Never reuse a gasket that has been fully torqued. The knife edges work-harden and deform the copper permanently.
  • If a flange was slightly loose, inspect the gasket for uneven bite marks — replace if in doubt.
  • Nickel-plated copper gaskets exist for higher corrosion resistance; use standard OFHC for baking to 400°C+.
  • Store gaskets in sealed bags — oxidised copper is harder and seals less reliably.

What never goes inside a UHV chamber

  • Zinc or cadmium — vapor pressure too high; will plate out on optics and destroy the vacuum.
  • Regular brass — contains zinc; dezincifies under vacuum and outgasses.
  • PVC, regular rubber, silicone (RTV) — extreme outgassing.
  • Fingerprints — the oils from your skin outgas for weeks. Always use nitrile gloves inside the chamber and clean parts with methanol + acetone before installation.
  • Unprocessed plastics — PEEK and PTFE are tolerable in small amounts; avoid everything else.

06 Reading Your Vacuum

Gauges, RGA, and the art of finding invisible leaks

Gauge types

GaugePressure rangePrincipleNotes
Pirani 10³ – 10⁻³ mbar Thermal conductivity of gas (heated wire) Gas-species dependent; great for rough/medium vacuum readout
Penning (cold cathode) 10⁻² – 10⁻⁹ mbar Ion current in crossed E+B fields No hot filament = no contamination; reads falsely high in strong B-fields
Bayard-Alpert (hot-ion) 10⁻² – 10⁻¹¹ mbar Electron emission → gas ionization → ion current collected Gold-plated collector reduces X-ray limit; standard UHV gauge
Ion pump current 10⁻⁵ – 10⁻¹¹ mbar I ∝ P × pumping speed Free readout from your existing ion pump supply; calibrate vs gauge
Spinning rotor gauge 10⁻² – 10⁻⁷ mbar Drag on magnetically suspended rotor Absolute, gas-independent; excellent for calibration

Residual Gas Analyzer (RGA)

Reading the mass spectrum

An RGA is a miniature quadrupole mass spectrometer that ionizes residual gas and sorts ions by mass-to-charge ratio. The peak heights tell you exactly what's in your vacuum. Key signatures:

  • m/z = 2 (H₂): Dominant peak in UHV — always. H₂ permeates stainless steel from the bulk. Normal, expected.
  • m/z = 12, 16, 28, 44 (C, O, CO, CO₂): Hydrocarbon contamination or outgassing from surfaces. Drop significantly after baking.
  • m/z = 18 (H₂O): High before bakeout, should be much smaller than H₂ after. If water stays high post-bake, suspect a small leak.
  • m/z = 14, 28 (N, N₂): Elevated N₂ usually means a real air leak. N₂/Ar ratio ≈ 3.7 confirms atmospheric air.
  • m/z = 40 (Ar): Unambiguous real leak indicator — Ar is not outgassed from steel.

Helium Leak Detection

Finding leaks down to 10⁻¹⁰ mbar·L/s

The He leak detector is a mass spectrometer tuned to m/z = 4, connected to your system's roughing line or directly to the pump port. He is small (0.26 nm kinetic diameter), inert, and rare in lab air (~5 ppm) — it threads through any crack the atmosphere leaks through, but faster.

Procedure:

  • Pump your system to the best vacuum you can achieve.
  • Spray He gas gently with a wand (or medical-grade He tubing) around suspect joints — CF flanges, viewports, electrical feedthroughs.
  • Wait 5–10 s (He transit time through a small leak).
  • A rise in the He signal identifies the leaking joint.
  • Start from the top and work down so He (lighter than air) doesn't drift upward into non-leaking regions.
💡 Sensitivity: ≈ 10⁻¹⁰ mbar·L/s. A "virtual leak" (trapped volume slowly releasing gas) mimics a real leak but won't respond to He spraying — the key diagnostic distinction.

07 Practical Checklist

From assembly to first UHV — a condensed field guide

Assembly phase
  • 1
    Clean all parts with acetone then methanol; handle only with nitrile gloves from this point forward.
  • 2
    Verify all materials: no zinc, cadmium, PVC, or standard rubber inside the system.
  • 3
    Use fresh OFHC Cu gaskets on every CF flange. Torque in star pattern: 3 passes from 3 → 8 → 14 N·m.
  • 4
    Install a gate valve between chamber and turbo pump. Verify it closes cleanly.
  • 5
    Install an ion gauge on the chamber. Label its filament voltage and emission current settings.
  • 6
    Leak check all connections with a He leak detector at rough vacuum (10⁻² mbar) before proceeding.
Pumpdown phase
  • 7
    Start scroll/diaphragm pump. Rough pump to <5×10⁻² mbar before starting the turbo.
  • 8
    Bring turbo to full speed. Open gate valve slowly. Monitor ion gauge.
  • 9
    Once below 10⁻⁶ mbar, begin bakeout ramp: <5°C/min up to 150–200°C.
  • 10
    Bake for 24–72h. Run ion gauge intermittently. Monitor RGA: water peak should decrease.
  • 11
    Cool at <3°C/min. Below 50°C: close gate valve, start ion pump, turn off turbo.
  • 12
    Fire TSP once (if installed). Ion pump current should read <1 nA for a 20 L/s pump at UHV.
You're at UHV when: Ion gauge reads <10⁻⁹ mbar, ion pump current is stable and <10 nA, RGA shows H₂ dominant with no N₂/Ar signature, and your pilot atom lifetime exceeds several minutes.
⚠️ Common failure modes: Virtual leak from trapped volumes (blind tapped holes, undrilled screws); fingerprint contamination on optics; bad CF torque — either sequence or final value; or baking too fast and cracking a ceramic feedthrough.
StageStart P (mbar)Target P (mbar)TimeKey action
Rough pump only1013~10⁻²10–30 minVerify no gross leaks
Turbo pump on10⁻²~10⁻⁶1–4 hoursHe leak test
Bakeout ramp10⁻⁶~10⁻⁴ (spike)2–5 hoursRemove heat-sensitive parts
Bakeout hold10⁻⁴~10⁻⁷24–72 hoursMonitor RGA, run turbo
Cool + ion pump10⁻⁷~10⁻¹⁰12–24 hoursTSP fire, NEG activate
See also: The vacuum system is just the starting point. Once you're at UHV, the next steps are trapping atoms with MOT & Magnetic Trap design, characterizing your imaging with the Imaging SNR Calculator, measuring the resulting atom temperature via TOF Thermometry, and transferring atoms to tweezers for single-atom thermometry.

See Also