What is a photomask? How is it made, and why does a set cost so much?

A photomask (also called a mask or a reticle) is the "negative" used for exposure: a transparent quartz plate carrying a circuit pattern. Where there's a thin metal film, light is blocked; where there isn't, light passes. The scanner shines light through it, then uses lenses to shrink the pattern and print it onto the wafer. Every layer of a chip's pattern needs at least one mask, and a full set often has dozens. Each one must be extremely fine and accurate, with not a single defect left, which is why they're expensive.

What does a photomask do? It works like a stamp

Chip patterns are printed with light. The step is called photolithography, or litho for short: the wafer is coated with light-sensitive photoresist, light shines through a photomask, the resist changes where the light hits it, and washing it leaves the pattern behind. The whole sequence is in What is photolithography?.

The pattern on the mask is usually drawn four times larger than on the chip, and the lenses shrink it before printing. The scanner prints one small area at a time, called an exposure field; after each field the wafer moves a little and the next one is printed, like using the same stamp to fill a whole sheet of paper. So one extra spot on the mask means one extra spot in the same place in every field, and it keeps printing, lot after lot.

What does a photomask look like?

• A very flat quartz plate: usually a square about 15 cm (6 inches) on a side and just over 6 mm thick. Quartz is a very pure glass that the ultraviolet light used for exposure can pass through.
• A light-blocking metal film: commonly chromium, shaped into the circuit.
• A pellicle: it floats above the pattern and keeps dust off it (more in step 5).

The most advanced scanners, which use extreme ultraviolet (EUV) light, are the exception: the mask itself is a mirror that reflects the light, with a light-absorbing pattern on top. See What is extreme ultraviolet (EUV) light?.

Why does one chip need dozens of masks?

A chip is built up layer by layer, like a building: transistors, the switches, at the bottom, and layer after layer of metal wiring above. Every layer has a different shape, so each needs at least one mask; the finest layers are sometimes split into two or three separate exposures, each with its own mask. That's why a full set often has dozens, and more for more advanced processes.

How is a photomask made?

Masks are made in dedicated factories called mask shops. Some belong to the fabs themselves; others make masks for anyone. The method is a lot like litho on a wafer: coat with resist, expose, develop, etch. The difference is the exposure step: a mask is the very first negative, so there's no other negative to print it from. It has to be drawn, stroke by stroke.

Step 1: Turn the design file into something that prints

• What: When a chip's design is finished and handed off, that's called tape-out. The patterns of the finest layers are first corrected by computer, which is called OPC (optical proximity correction): the shapes are deliberately drawn a little differently from what's wanted, such as a small extra block at the end of a line. The corrected pattern goes to the mask shop, where it is scaled up four times and cut into small pieces a machine can write.
• Why: The lines are about as fine as the wavelength of the exposure light, or finer. Printed as drawn, corners come out rounded and line ends come out short.
• Otherwise: What prints doesn't match the design, and things that should connect don't. On OPC, see What is OPC?.

Step 2: Write it on, stroke by stroke, with an electron beam or a laser

• What: Start with a blank mask: a quartz plate already coated with chromium and then with a layer of resist. The mask writer follows the file with a tightly focused electron beam, exposing it one small piece after another; layers that aren't as fine can be written with a laser instead.
• Why: An electron beam can be focused extremely finely and can draw the finest patterns; a laser is faster but can't draw as fine.
• Otherwise: If the machine drifts even slightly while writing, the pattern it draws lands in the wrong place; printed on the wafer, that layer won't line up with the one below (see What is overlay?).

Step 3: Develop and etch the pattern into the chromium

• What: Develop washes away the exposed resist (some resists work the other way round), uncovering the chromium underneath; the uncovered chromium is etched away, and the remaining resist is removed.
• Why: It's the chromium that actually blocks the light; the resist is only a stencil along the way.
• Otherwise: If the chromium isn't fully etched, light is still blocked where it should pass; etch too far, and the lines get thinner.

Step 4: Measure, inspect, repair

• What: Measure how wide the lines are and whether they're in the right place. An inspection tool scans the whole plate, comparing two identical patterns with each other or with the design file, to find extra bits, missing bits and specks. Defects it finds are fixed with a very fine electron or ion beam, removing what's extra and filling what's missing.
• Why: One printable defect on a mask prints in every field of every wafer.
• Otherwise: One missed spot can ruin many chips at once, and yield drops. A mask that can't be repaired has to be made again.

Step 5: Clean it and mount the pellicle

• What: A final clean and one more inspection, then the pellicle goes on: a frame holding a very thin transparent film a few millimeters above the pattern.
• Why: Dust that lands on the film is too far from the pattern to be in focus, so it usually doesn't print.
• Otherwise: Dust that lands right on the pattern prints sharply in every field. How pellicles are looked after later is in What does a mask engineer do in a chip fab?.

Step 6: Box it and ship it to the fab

• What: The mask goes into a dedicated mask pod with its inspection report and is shipped to the fab.
• Why: On the way it can't be touched, can't pick up dust and can't get damp.
• Otherwise: A particle that gets under the pellicle on the way and lands on the pattern prints just the same (one on the film usually doesn't); so the fab inspects the mask again when it arrives.

If an electron beam can draw it, why not draw straight on the wafer?

Because it's too slow. A mask only has to be drawn stroke by stroke once; after that the scanner just flashes each field, so even a wafer with dozens to a hundred or more fields is printed quickly, and the same mask can print a great many wafers. Drawing every wafer directly with an electron beam couldn't keep up with volume production, so electron beams are used mostly for making masks, for research, and for things made in very small numbers.

Why does a mask set cost so much?

• It has to be extremely accurate: the pattern shrinks four times, and so do its errors; but the finest lines on a chip are only tens of nanometers wide, so even the little error that's left can push a line width or a position out of spec.
• Not a single defect can stay: the whole plate has to be inspected and repaired without missing anything, and inspection alone takes up a large share of the time it takes to make a mask.
• Writing takes a long time: the more corrections, the more corners and edges and the more strokes to write, and usually the longer it takes. Mask writers and inspection tools are expensive, and the longer a mask ties them up, the more it costs.
• Many masks per set, and only one use: each mask is made for one layer of one chip, and no other chip can use it.
• More advanced means more expensive: finer lines, more corrections, more masks, and EUV masks are harder still to make.

How that money is spread over every chip is in Why are chips so expensive?. It's also why finding a design error after tape-out hurts so much: remaking the masks costs both money and time.

In fab24hr

The game doesn't show masks being made; the mask shop appears only on the phone. In the prologue of the main story "The life of a chip", Glowmere's design goes to the fab and is made into a set of masks, which is called "tape-out". In chapter 3, "The shape of a failure", you're the R&D integration engineer: on one wafer map, every exposure field fails in the same place, a problem that follows the mask; and at the 16:00 mask meeting you decide whether the next mask set keeps two spots for your own test keys.

To look after masks, tap "Clock in", choose "A day in the fab", then on the "Your role today" card tap "Production fab" and swipe to the mask engineer (marked Beta). Once a mask arrives at the fab, it's yours: in the morning, a mask comes back from cleaning and you do your own receiving inspection; in the afternoon, a pellicle has a scratch and that layer has no spare; in the evening, every exposure field has the same spot in the same place. The companies, the mask shop, the chip, the people, mask IDs and figures in the game are fictional; how masks are made and why they're expensive follow public, general sources. For other jargon, see Chip fab jargon in plain English.