What does MO mean in a chip fab?

On the floor of Taiwan's chip fabs, MO means mis-operation: a mistake a person makes while operating. Some also write it as Miss Operation. Running the wrong recipe, loading the wrong lot, picking the wrong wafers, running a lot at the wrong step, skipping a step: if a person got it wrong while operating, it's an MO. That's different from a tool breaking down on its own. A tool going down is an equipment problem; an MO is something that went wrong where people and the process meet. It's shop-floor shorthand, not a standard term, and the name and what counts as one vary from fab to fab. Research papers usually just say "human error", or say the wafers were "misprocessed".

Is the "Mo" in chip news the same thing?

Not necessarily. It depends on how it's written and where:

• Mo (capital M, small o): usually the chemical symbol for molybdenum, a metal. Some processes use it as a thin-film material. It has nothing to do with operating mistakes.
• The MO in MOCVD: it stands for metal-organic. MOCVD is a way of growing thin films, often used to make compound semiconductor devices such as LEDs.

What does an MO usually look like?

Two words first. A recipe is the set of settings a tool runs: which gases, what temperature, how many seconds. A lot is usually one box of 25 wafers. Common MOs look like this:

• Wrong recipe: two names on the list look alike, and the other one gets picked.
• Wrong lot: two boxes look the same, and the other lot goes on the tool.
• Wrong wafers: only a few wafers need the step, but the screen has the whole box selected.
• Wrong step: the lot isn't due for this step yet, or already had it and gets it again.
• Skipped step: a step that should happen doesn't, and the lot moves on.

Most of the time, the wafer numbers and recipe reach the tool from the system through a work order, so nobody picks them by hand. MOs are especially likely in operations outside the normal flow, where a person has to choose: a manual run, for example finishing a wafer that got stuck halfway when a tool went down. A paper on human error in wafer fabs makes the same point: after a tool interruption, people handling partly processed wafers by hand can misprocess them, and the potential loss is large. For what a recipe is, read What is a recipe in semiconductor manufacturing?.

Why do fabs fear MOs so much?

Three reasons:

1. A mistake often hits a whole lot: tools mostly work lot by lot, so one choice can apply to every wafer in the box.
2. Many steps can't be undone: a mistake in litho can usually be fixed by stripping the photoresist and doing it again, which is called rework (why that works: Why is the lithography area in a chip fab yellow?). Once etch has removed material or implant has driven atoms in, there's no taking it back, and the wafers often have to be scrapped.
3. The later, the costlier: a wafer goes through hundreds of steps over several weeks. A mistake near the end throws away all the time and material that went in before it.

Even with nothing scrapped, the records stop matching: the system says the instruction was followed, and the tool actually did something else. The lot has to stop while someone goes through it wafer by wafer for disposition, and the delivery date slips.

What's the first thing to do after an MO?

Once you notice the mistake, three things can't wait:

1. Stop the lot: put it on hold so it doesn't move to the next step. Every extra step a wrong lot takes is one more step to clean up (how lots normally move: What does the manufacturing department do in a chip fab?).
2. Tell people: tell whoever owns the lot and your manager right away. The sooner you speak up, the more can be saved.
3. Keep the scene as it is: the tool's run log, what was selected on the screen, the instruction at the time. Don't rush to change them. Change them now, and later no one can see what things looked like.

Causes and fixes come later. Right now, what matters most is laying out clearly what happened. Public workplace-safety guidance on incident investigations says the same: secure the scene and keep the evidence first, then look for causes.

Is the review about finding who's to blame?

No. A good review asks what in the process made the mistake easy, not who was careless.

People make mistakes, however careful they are. Safety research has a well-known picture for this, the "Swiss cheese model": every layer of defense is like a slice of cheese with holes in it, and a mistake gets all the way through only when the holes happen to line up. Look only at the last person, and the holes stay where they are; the next person will run into them too. The same safety guidance warns that writing the cause down as "carelessness" or "didn't follow the procedure" is too easy and often misleading, because it never finds what really needs to change.

Holes that reviews often find:

• Two options with names so alike you can't tell them apart at a glance.
• A screen whose default value isn't the one you need this time.
• Wafer numbers and recipes copied from a chat message, or kept in someone's head.
• So many warnings that people click them away out of habit.

A graduate thesis on the screens of a wafer fab's manufacturing system found the same thing: badly designed screens make operators more likely to slip and put them under more pressure.

The review goes into an incident report. A common format covers, in order: what happened, containment (checking first whether other lots were hit too), the root cause, and how to keep it from happening again.

How do you stop the same mistake from happening again?

Fixes come in roughly three kinds, the first the most reliable:

1. Let the system block it: make the wrong step impossible. For example, the system fills in the wafer numbers and recipe and the person only presses Run; before the tool starts, the system checks that this is the right lot and the right recipe, and won't run if they don't match. This kind of design is called mistake-proofing.
2. Two people check: one person operates, another checks against the instruction, and only then do they press Run. One person's slip gets caught on the spot by the other.
3. Reminders and training: add the warning to the SOP, put a note by the screen, teach it again. It helps, but it's the weakest, because it still depends on everyone remembering every single time.

Workplace safety ranks fixes the same way: removing a hazard by design beats relying on rules and reminders. The last step is horizontal deployment: once this hole is closed, check other tools and other sections for the same hole, and close those too.

In fab24hr

In fab24hr, MO is at the heart of the etch process engineer's afternoon shift. Tap "Clock in", choose "A day in the fab", then on the "Your role today" card tap "Production fab", swipe to the process engineer, and tap "Etch" under the name. Which of the four shifts you get follows the time you open the game; between 12:00 and 18:00 it's the afternoon shift. The wafer that got stuck in ETCH-03 overnight needs a 15-second rework etch, and you have to do it yourself. That kind of fix can only be done as a manual run at the tool: the screen opens with all 25 wafers already selected, the recipe is the one used last time, and the list still has two retired versions with names that look alike. Next to it is a button to ask Kai, the equipment engineer, to take a look.

After you press Run, the afternoon goes one of two ways. If you make the MO, you get Yun in Manufacturing to put the lot on hold, tell Ya-Ting in Integration which wafers ran what, and report to the chief, whose first words are "Understood. No need to apologize right now. Laying out clearly what happened matters more." If you don't, the MO happens next door on ETCH-05, to the new hire Xiao An, and you write the report with him. At 15:00 comes the incident report: apart from what happened, the other three columns have to come from three different people. At 16:30 comes the review meeting, and each of three countermeasures needs an owner. Chapter 8 of "The life of a chip", "In a Hurry", is this same afternoon. As the quality engineer or in production control, the afternoon shift also brings the manual run gone wrong on ETCH-05: one keeps the scene as it was, the other replans the week's shipments. The companies, people, tool numbers, wafer counts and screens in the game are fictional; how mistakes happen and how they're handled come from public, general sources; the name MO and real practice differ from company to company. For other jargon, see Chip fab jargon in plain English.