Vacuum Technology: The Complete Engineering Guide — Regimes, Pumps, Conductance, Outgassing, Leak Detection, and the Industrial Processes Vacuum Enables

Vacuum is the most invisible input in manufacturing. Nobody buys "vacuum" the way they buy steel or electricity — yet remove it and a startling fraction of modern industry stops within one cycle: every PVD-coated cutting tool, every vacuum-brazed heat exchanger and cold plate, every bright-hardened die and bearing, every semiconductor layer ever deposited, every satellite component qualified for space, and the thin-film coatings that make phone screens and solar panels possible. The pump market that supports all of this is one of the least glamorous and most critical slices of the industrial supply chain.

Here is the scale of what "vacuum" spans. The span from atmosphere to a good ultra-high vacuum is about fourteen orders of magnitude of pressure — from 1,013 mbar down to 10⁻¹¹ mbar. Put differently: the pressure range a routine vacuum system controls is wider than the fall from sea level (1,013 mbar) to low Earth orbit, where ambient pressure is roughly 10¹¹ times lower. And a vacuum is never empty. At 10⁻⁶ mbar — a perfectly respectable high vacuum — a cubic centimetre still contains about 25 billion molecules. At 10⁻⁹ mbar, UHV territory, you're down to about 25 million per cm³. The engineering problem is never "no gas"; it is always "which gas, how much, and how fast can I get it to leave".

This guide covers the whole stack the way a plant engineer actually meets it: the physics that constrains every design (Section 1–2), the pump families and how to choose between them (Section 3), system design with conductance and pump-down mathematics (Section 4), the water problem that actually limits your base pressure (Section 5), measurement and leak detection (Section 6), and then the processes — PVD coatings, vacuum heat treatment and brazing, electron beam welding, vacuum workholding, casting and forming (Sections 7–9) — followed by economics, maintenance, and the Indian vacuum industry, which is older and deeper than most engineers realise (Sections 10–11).

---

1. The Pressure Scale, and Why the Regime Rules Everything

Units, because this field refuses to standardise

Vacuum is the last corner of engineering where four unit systems coexist in the same datasheet. The table below is the one worth memorising:

Unit · Equals · Typical use

1 atm · 1,013.25 mbar = 760 Torr = 101,325 Pa · reference

1 mbar · 100 Pa = 0.750 Torr = 750 µmHg · SI-adjacent, most modern pump datasheets

1 Torr · 1.333 mbar = 1,000 µmHg ("micron") · older literature, US practice

1 µmHg ("micron") · 1.333 × 10⁻³ mbar · Indian and US heat-treatment shops ("10 micron leak-up")

When a vacuum furnace or EB welder spec says "10 micron/hour leak rate", it means a pressure rise of 10 µmHg per hour — 1.33 × 10⁻² mbar/h. When a sputtering paper says "0.3 Pa", that is 3 × 10⁻³ mbar. Half of all vacuum troubleshooting is unit conversion done correctly; the other half is remembering that gauge readings depend on which gas is present (Section 6).

The regime map

The industry divides the scale into regimes, and the exact boundaries vary by source — NIST, Pfeiffer, and the furnace manufacturers all draw the lines slightly differently. The operative version, with the processes that live in each band:

Regime · Pressure range · Molecular density · What lives here

Rough / low vacuum · 1013 → 1 mbar · 2.5 × 10¹⁹ → 2.5 × 10¹⁶ /cm³ · vacuum forming, vacuum bagging, degassing, vacuum drying, impregnation

Medium vacuum · 1 → 10⁻³ mbar · 2.5 × 10¹⁶ → 2.5 × 10¹³ · freeze drying, some degassing and distillation, roughing for every high-vacuum system

High vacuum (HV) · 10⁻³ → 10⁻⁷ mbar · 2.5 × 10¹³ → 2.5 × 10⁹ · PVD coatings, EB welding, vacuum brazing, vacuum hardening, electron microscopy

Ultra-high vacuum (UHV) · 10⁻⁷ → 10⁻¹² mbar · 2.5 × 10⁹ → 2.5 × 10⁴ · semiconductor surface processes, accelerators, surface science, gravity-wave detectors

The regime matters because the physics changes, not just the scale. In rough vacuum, gas is a continuum and pumps behave like compressors; a rotary vane pump moves litres per revolution, and leaks are audible hisses. In high vacuum, gas is a rarified molecular gas, and pumps move molecules — the design language switches to surface areas, conductances, and outgassing rates per square centimetre. In UHV, you are fighting the chamber itself: a single fingerprint is a contamination event, and every design decision is about walls, not the gas.

---

2. The Gas Physics a Vacuum Engineer Actually Uses

Mean free path: the single most useful number

The mean free path λ — the average distance a molecule travels between collisions — is what decides whether a pump "pulls" or "catches", whether a pipe passes gas or throttles it, and whether a sputtered atom reaches your part in a straight line:

For air at 20 °C (molecular diameter d ≈ 0.37 nm), this evaluates to:

The numbers that matter in practice:

Pressure · Mean free path (air, 20 °C)

1 atm (1013 mbar) · 68 nm

10 mbar · 6.8 µm

1 mbar · 68 µm

0.1 mbar · 0.68 mm

1 Pa (10⁻² mbar) · 6.8 mm

0.1 Pa · 68 mm

10⁻⁴ Pa (10⁻⁶ mbar) · ~68 m

At atmosphere, molecules collide every 68 nm; gas behaves as a dense crowd. At 1 Pa — a typical sputtering working pressure — λ is 6.8 mm: comparable to a feature on the part, so sputtered atoms fly nearly line-of-sight, but still collide enough to scatter ("thermalise") into a soft arrival distribution. At 10⁻⁶ mbar, λ is ~68 m: a molecule entering the chamber from a leak will, with overwhelming probability, hit a wall before hitting another gas molecule. In high vacuum, gas transport is a surface phenomenon: molecules bounce wall-to-wall until a pump catches them. That is exactly why the wall material, surface finish, and cleanliness of a chamber matter as much as the pump bolted to it.

The ratio of λ to the characteristic dimension d of the chamber or duct is the Knudsen number Kn = λ/d, and it defines three flow regimes:

• Viscous (continuum) flow, Kn < 0.01: gas behaves as a fluid. Pressure differences drive mass flow; Poiseuille's law rules; pumping is fast and intuitive.
• Transitional flow, Kn ≈ 0.01–1: the messy middle where most "why won't it pump down" problems live.
• Molecular flow, Kn > 1: molecules move independently; flow no longer depends on pressure difference but on molecular arrival rates. Conductance formulas change completely (Section 4), and so do pump curves.

The monolayer clock

The kinetic theory gives the impingement rate — molecules striking a surface per unit area per second:

For air at 20 °C, at 10⁻⁶ mbar, Z ≈ 2.9 × 10¹⁴ molecules/cm²·s. A "monolayer" of adsorbed gas is roughly 10¹⁵ sites/cm² (about one molecule per square nanometre). So at 10⁻⁶ mbar, an atomically clean surface is re-covered by a complete gas monolayer in about 3.5 seconds — even in a "good" high vacuum. At 10⁻⁹ mbar it takes about an hour.

This is the deepest design constraint in the field, and it explains behaviours that otherwise look like stubbornness:

• Why UHV costs so much more than HV. Going from 10⁻⁶ to 10⁻⁹ mbar is not "1000× better"; it's typically two more pump stages, all-metal seals, a bakeout system, and cleanroom discipline — because the acceptable wall flux drops 1000× while the wall area stays the same.
• Why PVD deposition must be fast. A tool coating deposited in 30 minutes at ~2 nm/s out-races the residual-gas arrival rate by orders of magnitude; the atoms that build the film arrive far faster than the contaminants can. Slow the deposition rate and the film chemistry degrades measurably.
• Why surface science is done at all. A copper crystal measured at 10⁻⁶ mbar is a copper crystal covered in yesterday's air by the time the experiment starts.

Gas composition: it's mostly water

A fresh chamber does not contain "air". It contains, in rough dominance for an unbaked system: water vapour (typically 60–95 % of the residual gas in a metal chamber within hours of venting), plus nitrogen, oxygen, hydrogen, argon, carbon oxides, and — inside any system with oil-sealed pumps or elastomer seals — hydrocarbon fragments. Hydrogen also diffuses through stainless steel walls at UHV (it outgases from the bulk, not just the surface). The water story has its own section (Section 5) because it is, quantitatively, the thing that decides your pump-down curve.

---

3. Pump Families: Base Pressure, Speed, Selectivity, Cost

Vacuum pumps split into gas-transfer pumps (they move gas mechanically to the outside, at exhaust pressure) and capture pumps (they trap molecules on surfaces inside). No single pump covers more than three or four decades of the fifteen-decade range, which is why every real system is a train: roughing stage → high-vacuum stage → sometimes UHV stage.

Family · Speed range · Base pressure class · Works from · Role · Watch-outs

Diaphragm · 0.5–10 m³/h · ~1–10 mbar · atmosphere · oil-free roughing, backing small turbos · low flow, limited compression

Rotary vane (two-stage) · 2–100 m³/h · ~10⁻³ mbar · atmosphere · the workhorse rough/backing pump · oil changes, backstreaming if foreline over-pressurised, water in oil with gas ballast off

Dry scroll · 3–50 m³/h · 0.05–0.3 mbar class · atmosphere · clean roughing, backing; quiet, no oil · tip-seal wear, price

Roots blower ("booster") · 100–10⁴+ m³/h · (backed: 10⁻³ class) · ~50 mbar · fast roughing of big chambers · needs a backing pump, high compression only at low Δp

Turbomolecular · 10–5,000 L/s · 10⁻⁷–10⁻¹⁰ mbar (baked) · ≤ 1 mbar · clean high vacuum: PVD, EBW, semis, R&D · bearing care, gas-dependent compression, sudden-vent damage

Diffusion · 100–50,000+ L/s · 10⁻⁶–10⁻⁹ mbar · ≤ 0.1–0.5 mbar · cheapest speed at scale, robustness; furnaces, coaters · oil backstreaming, warm-up, cold traps for clean work

Cryopump · 500–60,000 L/s · 10⁻⁷–10⁻¹¹ mbar · ≤ 10⁻² mbar · fast, clean UHV capture; semiconductors, space sim · finite capacity, regeneration, no helium pumping, vibration

Ion / getter (incl. NEG) · 1–500 L/s · 10⁻⁹–10⁻¹² mbar · ≤ 10⁻⁶ mbar · sealed systems, accelerators, UHV maintenance · start-pressure limited, noble-gas handling quirks

The turbomolecular pump, properly understood

A turbo is a multi-stage axial turbine spinning at 24,000–90,000 rpm (small rotors spin fastest), dragging molecules toward the exhaust with blade surfaces inclined so atoms preferentially acquire downward momentum. The physics is molecular flow, so a turbo can't start at atmosphere — a backing pump must first bring the chamber to ≤1 mbar, and the turbo then finds that pressure itself, backing through to atmosphere.

Two datasheet facts separate a real understanding from brochure-reading. First, compression ratio is wildly gas-dependent. From the Agilent TwisTorr 84 FS datasheet: compression ratio for N₂ ≥ 1 × 10¹¹, for helium 2 × 10⁶, for hydrogen 5 × 10⁴. The bigger 804 FS gets helium to 2 × 10⁸ and H₂ to 3 × 10⁶ via added "drag stages". Why care? Because the partial pressure at the chamber is set by (foreline partial pressure ÷ compression ratio). If your backing pump produces 10⁻³ mbar of hydrogen at the foreline, a turbo with H₂ compression of 10⁶ only holds the chamber hydrogen to ~10⁻⁹ mbar — often the floor of a baked UHV system. Light gases are the hard part of vacuum; hydrogen is the hardest of the common ones.

Second, base pressure specs are conditional. ISO 21360 defines how base pressure is measured: leak-free dome, 48 hours of bakeout, defined ambient. A turbo's catalog 10⁻¹⁰ mbar is real but earned; an unbaked chamber of the same pump will sit at 10⁻⁷–10⁻⁶ mbar and blame the pump. It's not the pump.

Why diffusion pumps still matter (especially in India)

The diffusion pump — invented by Irving Langmuir in 1915 — has no moving parts: a boiling oil jet entrains molecules downward in a staged stack. It is essentially obsolete in Western semiconductor fabs; it remains the dominant high-vacuum pump in cost-sensitive heat-treatment, coating, and metallurgical installations because nothing else provides thousands of litres per second for the price of a pump and a heater. Edwards, Pfeiffer, and Leybold all still catalog them; HHV (Section 11) manufactured them for decades. Their two liabilities are (1) backstreaming — hot oil vapour migrating toward the chamber, suppressed by cold traps, baffles, and correct foreline pressure; and (2) they can't pump against atmosphere, so a full roughing train is always underneath. For a vacuum furnace or a batch coater with 10⁻⁵–10⁻⁶ mbar requirements, a diffusion-pumped system is often the rational choice; for anything involving reactive or critical films, turbos win on cleanliness.

The dry revolution in roughing

Everything downstream of the chamber inherits the pump's exhaust chemistry. Traditional rotary vane pumps put oil mist and hydrocarbon fragments into the foreline; dry pumps (scroll, claw, screw) eliminate oil against the process gas, and became standard first in semiconductors and now in general industry: screw pumps for plasma/PVD/corrosive duty, claw pumps for food and packaging, scrolls for clean lab-scale work. In India's fabrication workshops the transition is slower — a vane pump plus an oil change schedule is durable, rebuildable, and locally supported — but any process with reactive gases (nitride coatings, CVD, plasma etch) will repay dry pumping with longer oil-free uptime.

---

4. System Design I: Conductance and Pump-Down Mathematics

The pump is never as fast as the number on the box

Gas at high vacuum is transported in two stages: the ductwork, then the pump. Each stage has a volumetric conductance C (L/s), and for series elements the effective pumping speed at the chamber is:

S_{\text{eff}} is always smaller than either contribution. This is the vacuum equivalent of a series resistor dropping voltage, and it is the most commonly ignored fact in the field: the useful pumping speed at your chamber is set by whichever of the pump and the plumbing is slower.

Conductance depends on flow regime. In the molecular-flow range, for a long circular tube of diameter d and length L:

and for a thin orifice, C \approx 11.6\,A L/s per cm² of area. Short ducts (L ≲ 2–3 d) get a correction: C \approx 12.1\,d^{3}/(L + 1.33\,d). Note the dependence: conductance scales with d³, so making a duct 2× wider improves its conductance 8×, while doubling its length halves it. Bellows, elbows, gate-valve bodies, and long elbows between chamber and pump can each cost you 30–60 % of your pumping speed. Every high-vacuum spec sheet that pairs a "fast" pump with a well-intentioned but narrow pumping port is quietly leaking money.

Worked example: a 1,200-litre coating chamber

Take a representative PVD batch chamber: V = 1,200 L (a 1.2 m diameter × 1.0 m cylindrical vessel, ~6 m² of wall plus fixtures), turbo pump with S = 1,000 L/s at the flange, connected through a DN 160 (16 cm) port and duct 50 cm long.

Step 1 — Hydrostatic roughing. From atmosphere to the turbo's working range. With a 30 m³/h (8.33 L/s) rotary vane pump, for a constant-volume-rate rough pump the pump-down equation gives:

from 1,013 mbar to 0.05 mbar. (Real time varies with the pump curve, condensables, and chamber cleanliness; 24 minutes is a fair mid-estimate for a clean system.) A 63 m³/h dry pump would cut it to ~11 minutes. Note that roughing is where most of the calendar time of a cycle goes for small chambers — optimise here first.

Step 2 — High-vacuum phase, ideal. DN 160 × 50 cm duct, long-tube formula gives C = 991 L/s, short-tube correction gives C ≈ 695 L/s. Then:

— you bought 1,000 L/s and delivered 410 L/s. A 2,000 L/s pump on the same duct would yield only 516 L/s; the duct, not the pump, is the bottleneck. From 0.05 mbar down to 10⁻⁶ mbar:

Step 3 — Reality check. No chamber reaches its vacuum in 32 seconds, because the pump-down equation assumes the chamber is empty of gas sources. The real curve flattens as outgassing from the walls and fixtures takes over — this is Section 5, and it is why a coating chamber takes 30–120 minutes to reach 10⁻⁶ mbar, and why an unbaked chamber can't reach 10⁻⁷ at all without heroic effort.

Practical design rules that follow

• Size the crossover deliberately. Turbo pumps start safely at ≤1 mbar, but each brand has a preferred crossover (often 0.05–0.5 mbar). Crossover too high: the turbo overheats compressing dense gas. Too low: wasted roughing time. Modern controllers automate this, but specify it explicitly.
• Put the pump where its speed can be used. A 1,000 L/s turbo behind a 25 cm ISO-K elbow assembly with a 90° bend and a bellows can see effective speeds below 300 L/s at the chamber. Where budget allows, mount the turbo gate valve directly on the chamber axis.
• Rough through a separate path. Big roughing lines and big HV lines have opposite requirements (viscous vs molecular flow); a single shared port compromises both. Typical architecture: separate roughing port with its own isolator, plus a dedicated HV port.
• Soft-start and soft-vent. Opening a high-vacuum valve instantly against a chamber at higher pressure slams the turbo and pushes oil vapour backwards from the foreline. Controller-managed soft-start valves and N₂ soft-vent (or full N₂ purge for reactive loads) are not options; they're protection for a ₹4–8 lakh pump.
• Account for the load, not the volume. For processes with gas flow (sputtering at 3 × 10⁻³ mbar argon, leak testing with helium spray), the pump must hold pressure against a throughput Q (mbar·L/s): p = Q/S_{\text{eff}}. PVD chambers are usually sized on throughput, not base pressure.

---

5. System Design II: Outgassing — the Water That Never Wants to Leave

Every surface in the chamber — walls, fixtures, seals, the parts themselves — is a reservoir of adsorbed and absorbed gas that slowly releases under vacuum. The steady-state base pressure of any system is set by the balance of gas load Q (mbar·L/s) against pumping speed:

Water dominates. Vent a chamber to humid air and the stainless steel wall adsorbs a multi-layer film of water. When pumped, that water desorbs slowly — this is why real pump-down curves (plot log p vs linear time) show a rapid initial fall then a long tail with slope ≈ −1 on a log-log plot, the classic signature of desorption-limited pumping. A rough hierarchy of outgassing rates for a stainless-steel chamber (per cm² of surface):

Surface condition · Outgassing rate (mbar·L/s·cm², order)

Freshly vented, warm, humid · 10⁻⁵ – 10⁻⁶

Pumped a few hours, room temperature · 10⁻⁷ – 10⁻⁸

Pumped overnight / days, unbaked · ~10⁻⁸ – 10⁻⁹

After 150 °C bake · ~10⁻⁹ – 10⁻¹⁰

After 250–300 °C bake (all-metal seals) · 10⁻¹⁰ – 10⁻¹¹

Feeding those numbers into the worked chamber above (internal wetted area ≈ 8 m² = 80,000 cm²; S_eff = 410 L/s):

Condition (q) · Gas load Q · Base pressure p = Q/S_eff

10⁻⁷ (humid, freshly vented) · 8 × 10⁻³ mbar·L/s · 2 × 10⁻⁵ mbar

10⁻⁸ (pumped hours) · 8 × 10⁻⁴ · 2 × 10⁻⁶ mbar

~3 × 10⁻⁹ (overnight) · 2.4 × 10⁻⁴ · ~6 × 10⁻⁷ mbar

10⁻⁹ (150 °C bake) · 8 × 10⁻⁵ · 2 × 10⁻⁷ mbar

10⁻¹⁰ (250 °C bake) · 8 × 10⁻⁶ · 2 × 10⁻⁸ mbar

Read that table once more, because it explains three things every vacuum engineer learns the hard way: (1) a pump-down that "stalls" at 10⁻⁶ mbar is not broken — it is outgassing-limited, and more pumping speed buys surprisingly little; (2) the only serious lever is the surface (cleanliness, bakeout, area); (3) base pressure improves by 10× for every decade you cut q, which is why UHV people spend their lives on surface preparation instead of pump shopping.

The actionable parts of the outgassing table

Clean the inside of the chamber like it's a cleanroom part. Machining oils, grease, and fingerprints are outgassing reservoirs of 10–1,000× the base surface rate. Standard practice: solvent wash (including a wipe-down with acetone/isopropanol… taking standard safety precautions) → ultrasonic clean where possible → deionised water rinse → dry. For extreme UHV, electropolished stainless with a controlled oxide layer (the famous CERN-style "stainless steel recipe") is the standard.

Bake what you can, at the highest temperature the components allow. Viton (FKM) O-rings limit bakeout to roughly 150 °C (200 °C short-term); perfluoroelastomer (FFKM, e.g., Kalrez) to ~250 °C; all-metal CF seals tolerate 250–400 °C. Heating jackets or heater tapes at 120–250 °C for 12–24 h, applied to chamber and especially to the pumping duct, routinely improve base pressure by 10–100×. Bake the gauges too — a cold gauge is a virtual leak.

Fixture discipline. In a coating or heat-treat chamber, half the surface area is tooling. Painted or anodised fixtures, adhesive tapes, and porous materials (foams, ceramics) hold water for days. Stainless, glass, ceramic, and aluminium fixtures bake clean; everything else is a delay.

Never vent with humid air if you can vent with dry nitrogen. A dry-N₂ vent turns a multi-hour water re-equilibration into minutes; the difference shows on every subsequent pump-down, forever.

The leak-rate test: the system's health certificate

The standard acceptance test for any vacuum vessel — furnace, chamber, or dewar — is the pressure-rise (leak-up) test: pump to a good vacuum, isolate the pumps, watch the pressure rise over 30–60 minutes. Any rise faster than the spec is gas ingress and/or outgassing. A tool-steel vacuum furnace typically accepts <10 µmHg/h (≈ 1.33 × 10⁻² mbar/h); for a 2,000 L vessel, that corresponds to a leak rate of:

— which, as Section 6 explains, is a very large leak in helium-leak-detector terms (a small O-ring scratch). Furnace specs are generous because furnaces are rough vacuum devices; a cryogenic or semiconductor vessel with a spec of 10⁻⁸ mbar·L/s is where leak detection actually becomes an art.

---

6. Measurement and Leak Detection

No single gauge covers the range

Vacuum gauges come in two philosophical families: direct gauges measure force (gas pressure acting on a diaphragm) and are gas-independent; indirect gauges measure a gas property (thermal conductivity, ionisation) and read differently for every gas. The standard toolkit:

Gauge · Range · Type · Notes

Thermocouple / Pirani (thermal) · 1000 → 10⁻³ (10⁻⁴) mbar · indirect · cheap, rugged, gas-species-dependent; the shop gauge

Capacitance manometer · 1000 → 10⁻⁵ mbar · direct · the calibration reference; accurate, gas-independent, expensive

Cold-cathode (Penning) · 10⁻² → 10⁻⁷ mbar · indirect · robust, no filament to burn, slight nonlinearity, needs a discharge strike

Hot-cathode ion (Bayard–Alpert) · 10⁻⁴ → 10⁻¹⁰ (10⁻¹¹) mbar · indirect · the UHV standard; sensitive to contamination; sensitivity ~10–25 mbar⁻¹

Spinning rotor gauge · 10⁻² → 10⁻⁷ mbar · transfer standard · used to calibrate other gauges

Full-range combination · 1000 → 10⁻⁹ mbar · mixed · e.g., Pirani + cold cathode in one body; what real systems mount

Two subtleties worth knowing. First, gas sensitivity: a gauge calibrated for nitrogen reads differently for argon (often 30–100 % error) and wildly differently for helium and hydrogen — so a gauge reading during a helium leak test is not a pressure measurement in the usual sense. Second, thermal transpiration: at very low pressures, gauges connected by small tubes read pressure differently at different temperatures — mostly a UHV calibration concern, but worth a line in any metrology procedure.

Leak detection: from soap bubbles to mass spectrometry

Leak testing escalates with the required sensitivity:

Pressure-rise testing (above): sensitivity depends on volume and time; a 1 × 10⁻⁶ mbar·L/s leak into a 50 L chamber raises pressure by only 0.1 µmHg/hour — invisible to the rise test, easily found by helium.

Helium mass-spectrometer leak detection is the industry's gold standard because helium is inert, small (passes through leaks a nitrogen molecule wouldn't), and nearly absent from air (~5 ppm — the background the detector must discriminate against). Two methods:

• Vacuum mode (spray): detector pumps the part/system, helium is sprayed on suspect joints from outside; sensitivity to 5 × 10⁻¹² atm·cc/s (≈ 5 × 10⁻¹² mbar·L/s) class for modern detectors. This is the "part escapes a few cm³ of gas in centuries" regime that leak-detector brochures love quoting.
• Sniffer mode: the part is pressurised with helium and the probe sniffs escaping gas; sensitivity ~1 × 10⁻⁷ mbar·L/s — 5 orders less sensitive, but no chamber needed and you can localise on a live line.

Practical acceptance thresholds, in mbar·L/s: hermetic microelectronic packages ~10⁻⁸–10⁻⁹; vacuum-brazed cold plates ≤10⁻⁸; diffusion-bonded ≥10⁻¹⁰ class (the tightest of the common joining processes); vacuum furnace heat zones acceptance per the Section 5 rise test (~10⁻² class). Calibrated helium leaks (typically 10⁻⁷–10⁻⁹, traceable) are the reference standard, and they should be recertified annually.

The diagnostic skill that matters most: distinguishing a real leak (a hole to atmosphere — pressure falls toward an asymptote and holds, rise test slope is constant and linear over time) from virtual leaks (trapped volumes: a threaded blind hole, stacked flanges, a bent gasket — gas ekes out for hours) and plain outgassing (pressure keeps falling slowly; the log-log slope is the desorption t⁻¹ fingerprint). Every vacuum engineer carries a ritual: if the curve hasn't plateaued, it's outgassing — be patient; if it flattens and holds, spray helium and find the hole; if the helium signal appears everywhere, check your own spray rig's "background" and zero the instrument before blaming the part.

---

7. PVD Coatings: The Highest-Volume Vacuum Process You Touch Every Day

Almost everyone in manufacturing touches PVD-coated items daily: coated end mills, drills, dies, punches, injection-mould gates, decorative watch cases, razor blades, cutting inserts. The vacuum environment exists because at 10⁻⁵ mbar or better, a freshly deposited film's contamination rate is low enough to grow a dense nitride or oxide with controlled stoichiometry.

The three industrial PVD routes

Route · Mechanism · Process pressure · Typical film rate · Signature characteristics

Evaporation (thermal / e-beam) · Heat source vaporises target; vapour condenses on substrate · 10⁻⁶–10⁻⁵ mbar · high (µm/min for e-beam) · line-of-sight, poor throwing power; optics, metallisation, thermal barriers

Sputtering (magnetron, esp. since ~1974) · Ar⁺ ions bombard a target; atoms ejected and deposited · 10⁻³–10⁻² mbar Ar · 1–10 nm/s · dense, smooth, excellent adhesion and composition control; the workhorse for functional films

Cathodic arc evaporation · High-current arc evaporates metal; a highly ionised plasma jet condenses on biased parts · ~10⁻² mbar class (reactive N₂ for nitrides) · µm/hour class · highest ionisation (~up to 80–100 %), best for tool nitrides; macro-droplets (roughness) mitigated by filtering/steering

Tool coatings: what the vacuum buys you

The canonical comparison is TiN versus TiAlN, both deposited by reactive arc or sputtering:

• TiN is the classic hard, gold-coloured coating: ~20–24 GPa hardness, but it oxidises from around 550 °C — the brittle TiO₂ that forms breaks down the coating in hot machining, eroding its advantage over uncoated tools in sustained dry cuts.
• TiAlN adds aluminium: on heating, an Al-rich, dense oxide bilayer forms that passivates further oxidation, letting coated tools survive significantly higher temperatures and enabling faster, drier machining. This is why TiAlN-family coatings (TiAlN, AlTiN, AlCrN) dominate modern cutting tool catalogues.
• DLC (diamond-like carbon) is a different animal: amorphous carbon film with a friction coefficient around 0.1 — the go-to for mould release, sliding pairs, and wear parts where lubricants are unwelcome.

PVD coatings are typically 1–5 µm thick (adding ~1–3 µm per side). At sputtering rates of 1–10 nm/s, a 3 µm coating takes 5–50 minutes of deposition — and the vacuum system has to hold base pressure and stoichiometry for the whole run. Deposition temperatures for tool-grade coatings are typically 400–600 °C: hot enough to stabilise the film, but limited by the substrate temper — which is why the vacuum furnace operates near the tool steel's tempering limits, and why specialised low-temperature coatings exist for fully hardened components.

Everything in Sections 4–6 applies with a vengeance here: a sputtering chamber with a water-loaded wall has an oxygen partial pressure in the process gas; oxygen in the gas means oxides in the film. That is why coating shops obsess over chamber cleanliness and why the "same recipe" runs differently after a weekend shutdown or after changing the fixture set.

---

8. Vacuum Metallurgy: Heat Treatment, Brazing

Vacuum heat treatment: bright, clean, distortion-controlled

Vacuum hardening replaces atmosphere and salt-bath furnaces with a simple statement: below ~10⁻⁴ mbar, there is essentially no oxygen to oxidise or decarburise the surface. The results: parts come out bright and clean (no scale, no post-cleaning for many applications), distortion is reduced and reproducible, and there is no salt disposal problem.

The recipe level numbers:

• The furnace is evacuated to 10⁻⁴–10⁻⁵ mbar before heating — this removes the air and locks in surface cleanliness.
• At the austenitising temperature (e.g., 1,020 °C for H13 tool steel), the vacuum is deliberately backed off — a partial pressure of nitrogen — commonly 10⁻¹ to 100 mbar — because at high temperature and high vacuum, alloying elements (Cr, Mn, Co) begin to sublimate from the steel surface. This is the single most counter-intuitive fact for newcomers: heat treating at the best possible vacuum is wrong; the partial pressure is a process parameter.
• Quenching uses high-pressure gas (6–20 bar N₂; He and Ar blends for faster cooling): equipment nominally reaches cooling rates up to ~100 °C/min with helium. 10 bar N₂ gives roughly the severity of a moderate agitated-oil quench (quench intensity H ≈ 0.4), 20 bar N₂ approaches H ≈ 0.7–0.8.
• A typical H13 cycle: 1,020 °C austenitise → 6–10 bar N₂ quench → double temper at 540–600 °C → 44–52 HRC, with dimensional change of 0.05–0.15 % linear versus 0.2–0.5 % (and less predictable) for oil quenching. M2 high-speed steel gets a triple temper because its secondary hardening (M₂C/MC carbide precipitation) peaks around 540–560 °C.
• A representative mid-range furnace spec sheet: 600 × 600 × 900 mm work zone, 1320 °C max, ±5 °C uniformity (AMS 2750E), 5 × 10⁻⁵ mbar ultimate, 120 kW, 10 bar quench, ~100 °C/min cooling with helium, PLC + recipe control.

The health metric that ties back to Section 5: leak-up rate. A tool-steel furnace typically accepts <10 µmHg/h rise. And with ≤10⁻² mbar·L/s-scale leak acceptance, a slow air ingress during the 1,000+ °C soak means exactly the oxidation and decarburisation the process was bought to avoid. This is why furnace maintenance people talk about the leak test the way foundry people talk about melt chemistry.

(For the metallurgy of what happens inside those cycles — solution treatment, secondary hardening, tempering responses — see the site's complete guide to heat treatment of steel.)

Vacuum brazing: joining without flux, inside heat exchangers you use daily

Vacuum brazing uses a filler alloy that melts at 500–1,200 °C while the parent metal stays solid, and capillary action pulls the molten filler through the joint. The vacuum removes every gas that could interfere — no flux is used or needed, so there's no residue to clean and no flux inclusion. The pressure levels are material-specific:

• Stainless steel and nickel alloys: brace at 10⁻³–10⁻⁵ mbar; Ni-based fillers (e.g., BNi-2) typically run without partial pressure, with a preheat soak around 920–980 °C for uniformity.
• Copper and its alloys: heated to 1,100–1,120 °C under a partial pressure of inert gas up to ~1 mbar — the partial pressure restrains the evaporation of copper, the same sublimation logic as heat treatment.
• Aluminium: the demanding case — vacuum in the 10⁻⁵ mbar range or better at 575–590 °C with ±5.5 °C uniformity, using Al-Si filler (e.g., 4047). The window is tight because aluminium oxide is stubborn and the brace temperature is close to the alloy's own limits.

Where this shows up is everywhere in thermal hardware: plate-fin and bar-plate heat exchangers, cold plates for power electronics, rocket engine and satellite components, EV battery cooling plates, cutting-tool assemblies, hermetic sensor packages. A vacuum-brazed aluminium cold plate with internal channels down to ~0.5 mm is the workhorse of data-centre and inverter cooling, and the hermeticity offered by the best vacuum-joined structures is genuinely extreme — brazed assemblies are typically qualified to ≤10⁻⁸ mbar·L/s and diffusion-bonded (a cousin process, with pressure as well as heat) down to ≤10⁻¹⁰ mbar·L/s. If you've seen a FabFlow-style request for a "water-cooled cold plate" or conformal cooling mould insert, the vacuum brazing furnace is very likely in its future (here's the conformal cooling guide for that side of the story).

---

9. Electron Beam Welding, and Vacuum in Everyday Fabrication

EB welding: deep, narrow, and pressure-sensitive

An electron beam welder accelerates electrons through 60–150 kV and focuses them magnetically; the beam's energy density is so high that it vapourises a keyhole and penetrates tens of millimetres in a single pass (the AWS literature includes a 44.5 mm single-pass penetration in copper). But the beam must be in vacuum — the electron gun especially, where the rule of thumb is <10⁻⁴ mbar at the gun to prevent arc-overs, while the weld chamber runs in the high vacuum range, typically 10⁻³–10⁻⁶ hPa, with a forevacuum train (Roots + vane stations) crossing over at 10⁻¹–10⁻² hPa.

The subtle part, and a genuinely surprising result for anyone who assumes "more vacuum = better": beam behaviour is non-monotonic with pressure. In a controlled study, minimum weld width and maximum penetration occurred at an intermediate level around 10⁻⁴ Torr (≈1.3 × 10⁻⁴ mbar): at higher pressures the beam scatters; at lower pressures it loses space-charge neutralisation from ionised gas and the focus shifts. Practical translation: for high-precision EBW, chamber pressure is a controlled process variable, not just "lower is better" — production specs define a pressure band, not a floor.

Vacuum where you didn't expect it: the fabrication shop

Even if you never run a coater or furnace, vacuum is already in your workshop:

• Vacuum workholding (CNC): held part force is simply F = \Delta p \cdot A. A 300 × 300 mm vacuum table at −0.8 bar differential (80 % of full vacuum — normal on venturi or vane-pump tables) clamps with 7,200 N ≈ 730 kgf (0.09 m² × 80 kPa Δp) — enough for light milling and full-depth routing of sheet, and it releases, seats, and re-clamps in seconds. Drop the seal quality and the differential decays; leaky fixtures halve the force. Works on any flat porous or gasketed surface; struggles only with coolant (MDF and sacrificial layers absorb it, water vaporises under the seal).
• Vacuum forming and vacuum bagging: here you don't need fine vacuum at all — you need differential. Thermoforming works at an absolute pressure of ~100–200 mbar, because the sheet is pushed by ~0.8–0.9 bar of atmosphere, which is the highest force single-stage vacuum can ever apply. That limit is why thermoforming is a thin-sheet process. Same logic in composite vacuum bagging (0.1–0.9 bar Δp compacting laminates; see the composites guide).
• Vacuum casting and degassing: silicone-mould casting and resin degassing live at ~1–10 mbar. The chamber's job is to bring gas bubbles to the surface before the polymer cures — bubble removal is a race against viscosity rise, so the trick is applying vacuum early, often with a slow-then-fast ramp to avoid a foam-over.
• Vacuum impregnation (VPI): motors, transformers, and castings are impregnated with resin under vacuum (and often pressure) to fill porosity — the process responsible for the service life of countless windings.
• Freeze drying: lyophilisation must run below the triple point of water (6.1 mbar at 0.01 °C) — typically 0.1–1 mbar — so the ice sublimes directly to vapour without melting. It is why your instant coffee and injectable pharmaceuticals exist, and the same sublimation physics (latent heat 2,834 kJ/kg) is also why freeze drying is one of the most energy-expensive preservation methods in industry.
• Semiconductor processing: every layer of every chip passes through vacuum systems measured in UHV class for surface prep and 10⁻³ mbar class for plasma deposition and etch — the EUV story happens six or seven orders of magnitude cleaner than typical tool coatings.

---

10. Economics, Maintenance, and Failure Modes

What vacuum really costs

Energy is usually the smallest recurring cost, per cycle:

• Roughing 1,200 L to 0.05 mbar: ~24 min at ~1.5 kW ≈ 0.6 kWh. Turbo phase at 0.5 kW for ~30 min ≈ 0.25 kWh. Total ≈0.85 kWh per pump-down ≈ ₹7 at ₹8/kWh — trivial per cycle. At ~4,800 cycles/year (2 shifts, 300 days), that's ~4,000 kWh ≈ ₹32,000/year of pump energy — real, but nowhere near the dominant cost.
• The dominant costs are capital spread (a turbo-based HV system is a ₹lakhs-to-crores investment depending on scale), maintenance (oil, seals, bearings — and downtime), and process gas (argon/nitrogen flows at working pressure during PVD and leak tests).
• The big operational lever is uptime: a chamber that takes 40 minutes instead of 15 to reach base pressure doesn't fail loudly; it silently eats a third of your runtime. Water-loaded walls (Section 5) are the usual culprit, and baking/cleanliness are the fix.

Maintenance that actually determines base pressure

Component · Service · Typical interval

Rotary vane pump · Oil + filter change (more often with condensable/corrosive gas loads, gas ballast use, and dirty duty) · 1,000–3,000 h

Dry scroll · Tip seal replacement · ~1–2 years continuous

Turbomolecular · Bearing service/overhaul; avoid banking at high speed, sudden vents · ~40,000 h class; 3–5 years continuous

Diffusion pump · Oil change + clean; foreline pressure check · 1–5 years, load-dependent

Cryopump · Regeneration (warm-up and purge of accumulated gas) · days–weeks, load-dependent

Gauges · Calibration (especially for process-critical or UHV reading) · annual

Helium leak detector · Calibrated leak re-verification, internal vacuum health · annual

Troubleshooting table

Symptom · Most likely cause · First action

Pump-down stalls at ~10⁻²–10⁻³ mbar · Water/condensables, gas ballast open, worn oil or vanes · Run gas ballast cyclically, then check oil; start bakeout; verify valve positions

Reaches 10⁻⁵ but never lower · Real or virtual leak; O-ring outgassing; gauge offset · Rise test; helium spray; check every threaded blind hole and stacked flange

Pressure rises while heating · Fixture/insulation outgassing; heater feedthrough leaks · Clean/change fixtures; leak-test at temperature (thermal leaks differ from cold ones)

Oil smell / oily residue in chamber · Foreline overpressure, backstreaming, cold trap exhausted · Check foreline pressure; refurbish traps; consider dry roughing

Turbo "rumbles" or trips · Bearing wear, or gas load spike at crossover · Don't restart blindly; inspect, service bearings

Helium detector alarms everywhere · Background contamination (or your own spray) · Zero instrument, ventilate, recalibrate against a calibrated leak

Base pressure worse on humid days · Water adsorption on walls; humidity in the vent gas · Vent with dry N₂; bake after every exposure

---

11. India's Vacuum Stack — and Where FabFlow Fits

India's vacuum industry is 61 years old, and its anchor institution is Hind High Vacuum (HHV) in Bengaluru, founded in 1965. Its history is a walk through what a country builds when it decides vacuum is strategic infrastructure:

• 1967: India's first indigenous 12-inch laboratory vacuum coating unit, developed with the Indian Institute of Science.
• 1968: India's first high-vacuum rotary pump.
• 1970: an indigenous space simulation chamber to leak-test rocket thruster components.
• 1996: India's largest vacuum heat treatment and brazing furnace — 1.8 m diameter × 2 m hot zone, achieving 10⁻⁶ mbar.
• 2002: a vertical coating system that coated the 2.2 m mirror for the Himalayan telescope site at Hanle, Ladakh — at 14,800 ft above sea level.
• 2005: a robot-controlled sputter coating system for conductive/transparent coatings on aircraft canopies.
• 2008: the Horton-sphere vacuum installation at Trivandrum — 16.13 m diameter spheres with a pumping system of 120,000 m³/h cycle speed, coupled to a hypersonic wind tunnel.
• 2009–2011: the HHV Pumps joint venture (rotary, Roots, piston and combination pumps), exports of thin-film equipment to the UK, a 3.7 m-diameter telescope mirror sputter coater, and rotary vacuum brazing furnaces (3.65 m diameter) for brazing launch-vehicle engine nozzles.

Around HHV, the ecosystem includes the Indian service arms of Edwards, Pfeiffer, Leybold, Agilent and Busch (vacuum pumps are one of the truly global oligopolies still owned by a handful of firms), a growing pool of Indian vacuum system integrators, and the heat-treaters and coaters who run the furnaces: job shops in Pune, Coimbatore, Rajkot, Bengaluru doing vacuum hardening, brazing and PVD coating on a work-order basis. If you need a hardened punch, a brazed cold plate or a coated form tool in India, the capability usually exists within one state.

Where this meets fabrication: the components that make vacuum systems work are, to a machine shop, ordinary parts with slightly unusual requirements — chamber bodies and port components in machined aluminium and stainless, vacuum flanges (KF/ISO-K/CF families), feedthrough housings, baffle plates, cold-trap bodies, heat-shield brackets, valve bodies, fixtures for coating racks and brazing stacks. These are turned parts, milled flanges, laser-cut brackets and sheet-formed shrouds — normal work for the shops on FabFlow — the same way the CNC-machined parts that keep every other industrial system running get sourced. And on the demand side: if you need a cold plate, a vacuum-brazed assembly, or a coated/lockwired hardware set, those are exactly the kind of jobs FabFlow is built to quote. Post a drawing once; the vacuum-capable shops are already on the platform.

---

The One-Page Checklist

• Learn the units cold: 1 mbar = 100 Pa = 0.75 Torr; 1 µmHg = 1.33 × 10⁻³ mbar. Convert every spec you read before comparing it.
• Mean free path rules the design: λ ≈ 6.7 × 10⁻³/p[Pa] m — 68 nm at atmosphere, 68 µm at 1 mbar, 6.8 mm at 1 Pa. Know λ relative to your chamber dimensions to know your flow regime.
• The monolayer clock: ~3.5 s per monolayer at 10⁻⁶ mbar; ~1 h at 10⁻⁹ mbar. Deposition must be fast; UHV surfaces must be clean.
• Effective pumping speed, always: 1/S_eff = 1/S_pump + 1/C. Fix ducting (d³ × 1/L) before buying a bigger pump.
• Roughing takes the calendar time: t = (V/S)·ln(p₁/p₂); 1,200 L from atmosphere in ~24 min with a 30 m³/h pump. Then the HV phase is seconds in theory, hours in reality — because of water.
• Base pressure = Q/S_eff. It's an outgassing problem, not a pump problem. Clean, bake (150–250 °C as components allow), vent with dry N₂.
• Specify gas selectivity: turbos compress N₂ at ~10¹¹ but helium at ~10⁶ and H₂ at ~10⁴–10⁶ — the light gases set your UHV floor.
• Measure with the right gauge (Pirani for rugged shop duty; capacitance manometer as reference; Bayard–Alpert for UHV). All indirect gauges are gas-dependent — correct for helium and argon.
• Leak-test like a process: rise test for vessel acceptance; helium mass spectrometry (to 5 × 10⁻¹² mbar·L/s class) when it matters; distinguish leaks from virtual leaks from outgassing by the shape of the curve.
• PVD in one line: base ≤10⁻⁵ mbar, work at 10⁻³–10⁻² mbar sputter / ~10⁻² mbar arc; films 1–5 µm; TiN oxidises ~550 °C, TiAlN holds much further; cold base = bad film.
• Heat treat in one line: evacuate to 10⁻⁴–10⁻⁵, then set a partial pressure (10⁻¹–100 mbar) at temperature to stop alloy sublimation; quench 6–20 bar gas; leak-up <10 µmHg/h keeps you honest.
• Brazing in one line: stainless/Ni at 10⁻³–10⁻⁵ mbar, Al at ≤10⁻⁵ mbar at 575–590 °C; flux-free, hermetic to 10⁻⁸ mbar·L/s — the backbone of cold plates and heat exchangers.
• Respect the machine: ducts, soft-start, and dry-N₂ venting protect turbo pumps better than any warranty.

Vacuum is invisible precisely because it works. The pumps, gauges, and sealed rooms you never think about are the reason your coated tool lasts, your heat exchanger seals, and your sensors survive orbit. None of the physics in this guide will change — molecules will keep bouncing off walls at ~460 m/s, and water will keep being the most stubborn tenant in every chamber. What changes is whether your system is designed around that physics — or fights it every pump-down.