What is MES, and how does a chip fab know where every lot of wafers is?

MES stands for manufacturing execution system. In a chip fab, MES is the computer system that keeps track of every lot of wafers: what the lot is called, the hundreds of steps on its route, which step it is at now and where it goes next. When people in the fab talk about dispatching a lot, tracking it in or moving it out, it all goes through MES.

Why does a fab need MES?

Think of sending a parcel: it gets a tracking number, and it's scanned at every stop along the way. MES is a bit like that, except it doesn't only record. It also tells each step what to do with the lot, and stops the wrong thing before it happens.

A fab needs this more than most places. Each lot goes through hundreds of steps over several months. Chips are built up layer by layer, and for every layer the wafers go back through lithography and etch again. And the wafers all look the same: standing at a tool, you can't tell whether a lot is on layer 3 or layer 30. So "what does this lot do next?" can't be left to anyone's memory.

Many kinds of factories use an MES. A public international reference model puts it in the middle layer: above it are the systems for orders and planning, which work in months, weeks and days; below it are the controls that run the tools, working in seconds or less; MES works in days, shifts, hours and minutes.

Follow one lot: what does MES do at each step?

Step 1: wafer start, and the lot gets its number and route

• What happens: at wafer start, MES gives the lot a lot number, records which wafers are in it, and attaches its route: hundreds of steps, each saying which kind of tool to use and which set of conditions to follow. Each wafer also has its own ID laser-marked near its edge.
• Why: from now on, only the route knows which step the lot is on (for what happens that day, see "What is a wafer start?").
• Without it: people would have to remember how far each lot has got; a skipped or repeated step might not show up until many steps later.

Step 2: at a step, dispatching decides who goes first

• What happens: there are often several lots waiting for the same kind of tool. The dispatch system reads the data in MES: where each lot is, how urgent it is, how much time it has left, and which tools can run right now. Then it follows its rules to pick the next lot for each tool.
• Why: there are fewer tools than lots, so the order decides how long each lot waits. Which lots come first is set by Production control's plan (see "What does production control do in a chip fab?"); for how a rush lot cuts in, see "What is a hot lot?".
• Without it: with out-of-date data, a lot can be sent to a tool that has already stopped, or a lot close to its time limit can keep getting pushed back.

Step 3: before going into the tool, check

Before giving a patient medicine, a nurse scans the patient's wristband and then the medicine, to confirm it's the right person and the right drug. Before a lot goes into a tool, MES does much the same:

• Does this lot belong here? By its route, is this step really its next one? A lot at the wrong step, or one that skipped a step, is stopped here.
• Can the tool run it? The tool has to be in a state that allows production: a tool in maintenance, down, or repaired but not yet through its qual run is not. A tool whose maintenance can't be put off any longer should stop for it first. The tool also has to have the set of conditions this step needs.
• Which recipe? The system brings it up from the route and issues a work order, so nobody has to pick or copy the wafer numbers and recipe by hand (see "What is a recipe in semiconductor manufacturing?"). When a person does have to choose, picking the wrong one is an MO (see "What does MO mean in a chip fab?").
• Has Q-time run out? It counts from the moment an earlier step finished, usually the one just before. If it has run out, the wafer surface may already have changed, and the lot can't simply carry on down its route (see "What is Q-time?").

Why check before, not after? Because a tool doesn't know it has been given the wrong lot. It will carry out the wrong conditions very precisely, and many steps can't be undone. Only when all four are right does the lot "track in": the tool reads the ID on the wafer pod, and only if it matches does it open the door and take the wafers in one by one.

Step 4: finished, the lot moves out and the step is recorded

• What happens: when the tool is done, MES records one line: which wafers, which tool (for a tool with several chambers, which chamber can often be traced too), which recipe, who ran it, when it went in and when it came out. If this step includes a measurement, the numbers are stored under the lot number too. Then the lot moves out.
• Why: everything after depends on this line. Any Q-time starts counting from this moment, and the measurements go onto control charts.
• Without it: miss one entry, and whoever looks later will think those wafers never touched that tool.

Step 5: if something is wrong, hold

• What happens: if a measurement falls outside its limits, Q-time runs out, or a tool has a problem partway through, the lot is put on hold in MES. The system stops dispatching it and it can't track in at the next step until someone decides what to do with it (see "What is a lot hold?").
• Why: every extra step a suspect lot takes is more work spent on wafers that may already be bad.
• Without it: if it's only mentioned out loud, someone on the next shift may put it on the next tool.

Step 6: on to the next step

• What happens: a lot with no problems follows its route to the next step. The transport system acts on the dispatch decision and sends an OHT to carry the pod there (see "How do wafers move around a fab?").
• Why: MES already knows which kind of tool the next step needs, so the pod doesn't wait for someone to work out where to go.
• Without it: when it's all down to people remembering and pushing carts, pods go to the wrong tool more easily.

The same few steps repeat hundreds of times.

How does MES connect to the other systems?

A fab has more than one system, and each watches something different:

System	What it watches
Transport	which pod the OHTs take where
Metrology	how thick a film is, how wide a line is, how well layers line up
SPC	measurements plotted on control charts, to see whether a process is drifting
FDC	the tool's sensor readings: pressure, temperature, gas flow
APC	how much to adjust the next lot, based on the measurements

What ties them together is the same lot number, the same tool and the same time. MES works like a master ledger: everyone matches their data against "which lot, on which tool, at what time". So when a number goes out of limits, the fab can tell which lots to stop: when a point on an SPC chart falls outside the limits, or FDC sees a tool behaving oddly, people follow the OCAP and put the suspect lots on hold in MES. For what each system does, see "What is an SPC control chart?", "What is FDC?" and "What is APC?"; for how measuring works, see "How does a chip fab know each step went right?".

Why are the records kept?

Because when something goes wrong, the fab has to be able to trace back. This is called traceability. The international standard for quality management says the same: when products need to be traceable, each one has to be identifiable, and the records needed to trace it have to be kept.

A made-up example: this morning, an etch tool's monitor wafer measured out of limits, and the last normal reading was at ten last night. The first thing to do is ask MES: since ten last night, which lots has this tool run, and where is each one now? Those lots all go on hold first, and then they're judged one by one. With the records, only those lots are stopped. Without them, the fab would have to stop everything from that whole stretch of time, or gamble that it's fine.

It works the other way too. If a few lots test unusually badly when they're finished, the fab can compare the tools they went through and find the one that only the bad lots passed through (see "What is root cause analysis?"). And a chip may only show a problem long after it leaves the factory; what has to be checked then is still those lines from back then. So every line has to be right, complete and kept long enough.

In fab24hr

Tap "Clock in", choose "A day in the fab", then on the "Your role today" card tap "Production fab" and swipe to Manufacturing; you look after the etch area. At 19:45 on the evening shift, an MES warning pops up: P-2E08 has 30 minutes of Q-time left, but the lot isn't on the handover sheet. You open its lot history and go line by line to find where its Q-time started and who left it on the stocker. Sometimes the dispatch system goes offline, and every tool's track-ins and move-outs are stuck; whether the lots already running should stop is your call. As the quality engineer, on the afternoon shift you compare what the tool actually ran with what MES recorded, one section at a time.

The company, people, lot numbers, tools and times in the game are fictional, the system screens are simplified mock-ups, and the game treats the dispatch system and MES as one system. How MES works is written from public, general sources; real systems and practices differ from company to company. For other jargon, see Chip fab jargon in plain English.