Why do semiconductor tools need a vacuum?
Semiconductor tools need a vacuum because ordinary air is packed with molecules: atoms and ions flying to the wafer get knocked off course on the way, oxygen and water vapor oxidize the surface and get into the films, and the kind of even plasma a fab needs won't light properly. So plasma etch tools, most deposition tools and ion implanters pump the air out of the chamber before they work on a wafer, which is called pumping down. "Vacuum" here doesn't mean nothing at all, just very little gas; how little depends on the process.
How crowded is air?
In ordinary air, a molecule travels less than a ten-thousandth of a millimeter, on average, before it bumps into another one. This distance is called the "mean free path." It's like a packed train station: you can't take half a step without bumping into someone.
Cut the pressure to a tenth and that distance grows about tenfold. At around a millionth of atmospheric pressure, a molecule flies about 5 centimeters on average; at around a billionth, it can go tens of meters, like crossing an empty sports field.
What is the vacuum for?
- Atoms and ions fly straight to the wafer: metal atoms knocked off the target in sputtering (see What is thin film deposition?) and ions sped up in an implanter (see What is ion implantation?) get knocked off course or scattered when they hit gas molecules on the way.
- Oxygen and water vapor stay out: in ordinary air, if every molecule that hits a freshly cleaned surface sticks, the surface is covered by a layer of molecules in a few billionths of a second. At around a billionth of atmospheric pressure, that takes about a second. Oxygen and water vapor that do get in oxidize metal films and make them worse.
- The plasma lights, and lights evenly: when molecules are packed tight, electrons hit the next one before they can speed up, so it takes a very high voltage to light a plasma, and once lit it tends to pinch into thin threads like tiny lightning bolts. At low pressure, an even glow is much easier to get (see What is plasma, and why do chip fabs etch with it?).
- Heat only goes where it's sent: with no air moving around, the wafer's temperature depends mainly on the heater or cooled stage it sits on. And because a vacuum barely carries heat, etch tools put helium behind the wafer to cool it (helium backside pressure).
How empty does each process need it?
| Process | Pressure, roughly, as a fraction of ordinary air | How far a molecule flies on average |
|---|---|---|
| Ordinary air | 1 | Less than a ten-thousandth of a millimeter |
| Low-pressure deposition in a furnace | One part in a few thousand to one part in a few hundred | A few hundredths to a few tenths of a millimeter |
| Plasma etch | A hundred-thousandth to a ten-thousandth, or lower | A few tenths of a millimeter to a few millimeters |
| Sputtering (metal films) | First down to about a ten-billionth, then argon is let in to a millionth to a hundred-thousandth | During deposition, millimeters to centimeters |
| Ion implant (the path the ions travel) | About a billionth, or lower | Tens of meters or more |
Not every tool needs a vacuum: coater/developers, most scanners, wet clean tanks and CMP all work at ordinary pressure. Extreme ultraviolet scanners are the exception: that light is absorbed even by air, so its whole path has to be pumped down (see What is extreme ultraviolet (EUV) light?).
How is a vacuum made?
No single pump can take a chamber all the way from ordinary air to a very empty space, so it's done in two stages:
- Roughing: dry pumps remove most of the air first. "Dry" means no oil where the gas is pumped, so no oil vapor can flow back and dirty the wafer.
- High vacuum: what's left goes to a high-vacuum pump next to the chamber. In a turbomolecular pump, rows of blades spin at tens of thousands of revolutions a minute and bat the molecules downward, but it can only push them on to the roughing pump behind it, not straight out to the air, so the two work in series. A cryopump instead freezes gas onto surfaces colder than minus 250 °C; once they're full, it has to warm up, let the gas go and cool down again, which is called regeneration.
How does a wafer get in and out of the vacuum? Step by step
Many vacuum tools are shaped like a flower: a transfer chamber in the middle that always stays under vacuum, a few process chambers around it, and the load ports for wafer pods at the front.
Step 1: Into the load lock, door closed
- What: a robot puts the wafer into a small room at the front of the tool, the load lock, and the outer door closes.
- Why: it works like a spacecraft's airlock. The two doors never open at the same time, so the inner side always stays under vacuum.
- Otherwise: if wafers went straight into the main chambers, every wafer swap would mean letting air in and pumping down from scratch.
Step 2: Pump down the load lock
- What: the load lock is pumped down to about the transfer chamber's pressure, and only then does the inner door open.
- Why: the room is small, so it pumps down fast, and it's the only part that goes back and forth between ordinary pressure and vacuum.
- Otherwise: open it too early and air and water vapor come along. Pump too hard and the rush of gas can stir up particles, and water vapor can suddenly condense into tiny droplets that leave particles when they dry.
Step 3: Into a process chamber
- What: a vacuum robot in the transfer chamber puts the wafer into a process chamber and closes the narrow door between them (the slit valve).
- Why: the main chambers stay under vacuum the whole time, so they never have to be vented for each wafer.
- Otherwise: every time a chamber is vented, water vapor coats the inside walls, and it takes a long time to pump down again.
Step 4: Pump a bit lower, then let the process gases in
- What: the chamber is first pumped below the process pressure, then the gases go in according to the recipe, and a valve holds the pressure at its set point.
- Why: making sure stray gas is low first means only the gases you want are left.
- Otherwise: leftover oxygen and water vapor join the reaction, and the film or the etched shape comes out different from the settings.
Step 5: Pump clean, take it out, back to ordinary pressure
- What: when the process is done, the chamber gas is pumped out first, and the wafer goes back to the load lock. The load lock is slowly filled with dry nitrogen until it's back to ordinary pressure, and only then does the outer door open.
- Why: many of the specialty gases used are toxic or corrosive and mustn't come out with the wafer; the nitrogen is dry and let in slowly, so it brings in no water vapor and stirs up no particles.
- Otherwise: leftover gas gets into the transfer chamber; fill with ordinary air and water vapor coats the walls, so the next pump-down is slow.
What happens when there's a leak, and how do you find it?
Every seam of a chamber is sealed with a rubber O-ring or a metal gasket. Where a seal isn't right, air keeps seeping in: the pressure won't come down; oxygen gets into the films and oxidizes metal layers; and leaked air can react with process gases into a fine powder that lands on wafers as particles.
There are two common ways to find a leak. One is a rate-of-rise test: close the chamber off, stop pumping and watch how the pressure rises. A straight-line climb usually means a leak; a rise that keeps slowing down is usually just water vapor still coming off the inside walls. The other is a leak check: spray a little helium on one joint at a time from the outside and watch the detector on the chamber; the joint where the reading jumps most is the leak. That's why after a chamber open for repairs, it always has to be pumped down and leak-checked before anything else (see What does an equipment engineer do in a chip fab?).
Where are the pumps and the exhaust?
The dry pumps for roughing and the exhaust treatment that burns or breaks down the waste gas mostly sit in the subfab below the cleanroom, near the spot under their tool. The pumps run 24 hours a day, and many rely on cooling water to carry their heat away. For how much electricity they use, see Why do semiconductor fabs use so much electricity?; for who looks after that floor, see What does a fab facilities engineer do?
In fab24hr
Tap "Clock in", choose "A day in the fab", then on the "Your role today" card tap "Production fab", swipe to the equipment engineer and tap "Thin film" under the name: you're the Thin Film equipment engineer (marked Beta). On the late-night shift, "The slowly rising vacuum," PVD-01's pressure creeps up all night; in the rate-of-rise test it climbs more and more slowly, so it isn't a leak: the cryopump is nearly full. Regenerate it now, or let the lot finish first? On the afternoon shift you change a target: vent, pump down, rate-of-rise test. Tap "Implant" under the name instead, and on the evening shift, "End Station Vacuum," it's again a cryopump that's nearly full; tap "Etch" and on the morning shift you open the chamber to replace the ESC, then leak-check it with helium, one joint at a time. In "The life of a chip," chapter 7, "Firefighting," you decide how long to pump down and bake during the qual run.
To see how a wafer gets in and out of the vacuum, open the cluster tool and tap "See it work"; the pumps and exhaust treatment downstairs are in the subfab.
The companies, people, tool IDs and numbers in the game are fictional; how vacuum works and how pumps and load locks are used come from public, general sources. For other jargon, see Chip fab jargon in plain English.
Where you'll meet it in the gameAs the Thin Film equipment engineer (production fab › equipment › thin film), on the late-night shift PVD-01's chamber pressure creeps up all night: is something leaking, or is the cryopump nearly full? On the afternoon shift you change a target: vent, pump down, rate-of-rise test.
In fab24hr, the chamber pressure creeps up all night, and you have to tell whether something is leaking or the pump is running out of steam.
Play as the Thin Film equipment engineer: the targets, the vacuum and the cryopump are yoursTeachers: this article can go with "fab24hr in the classroom" in class.
Cite this article"Why do semiconductor tools need a vacuum?" fab24hr, updated . https://fab24hr.com/en/blog/vacuum/
Free to use in class and for non-commercial teaching: on screen, in handouts, printed for students. Please credit fab24hr. Do not sell it or present it as your own work. Spot a mistake? Report it
More articles
What is a fab in semiconductors, and what does "going into the fab" mean?
How clean is a cleanroom? What Class 100 and ISO 5 mean
What is a bunny suit, and why do chip fab workers wear one?
Why do chip fabs run 24 hours a day?
- What is fab24hr?: the two ways to play, the four shifts, common questions, sources
- The life of a chip: the life of one chip: tape-out, pilot run, mass production, wafer test, packaging, shipping
- A day in the life of a fab engineer: who does what (PE, EE, PIE, manufacturing), and the day hour by hour
- Who works in a chip fab: technicians, Manufacturing, process, equipment, Facilities, integration, QA, production control, masks, R&D, the chief, the vendor, and test and assembly at the test-and-assembly house: each role's day
- Chip fab jargon in plain English: MO, Hold, Q-time, OCAP and the rest, one plain sentence each
- Semiconductors 101: a free course following the life of a chip: A beginner's semiconductor course that follows the life of a chip: illustrated notes, slides, readings, 3D machines and discussion questions, for self-study or a class
- Chip fab machines, taken apart: What 31 chip fab machines do (etch, scanner, CMP, furnace, packaging…), each as a 3D model to take apart
- fab24hr in the classroom: Semiconductor lesson plans: 20/40-minute plans, a lesson per chapter or process area (3D machines, the game, articles), discussion questions, a worksheet, FAQ
- About and press kit: trailer, gameplay clips, descriptions and key facts