Digital Twins for Semiconductor Manufacturing

At a glance

Time needed

About 24 hours, at your own pace

Level

Intermediate

Who it is for

Undergraduate STEM students in their second year or later (physics, engineering, chemistry or computer science)

Prerequisites

College algebra and a little Python and statistics. The optional readiness check in Module 0 shows any gaps, and the prep pack helps you close them.

Format and what you need

Online and self-paced; a web browser and about 4 hours a week. Every lab runs in the browser, with nothing to install. A free Google account is optional and is needed only for the Google Colab copies of the labs.

Badge

Digital Twins for Semiconductor Manufacturing digital badge

What you will be able to do

• Define a digital twin using the Digital Twin Consortium and NIST IR 8356 definitions, place a system on the common model, shadow and twin classification, and tell three levels of coupling apart: updating the twin from observations, supporting a person's decision, and applying a control action automatically
• Trace the data path of a fab tool from sensors and metrology to the twin, clean public fab data with missing values, and align time-stamped records from different sources by run and wafer
• Calibrate a Deal-Grove thermal oxidation model to simulated furnace measurements, give the uncertainty of the predicted mean and of a new measurement, and state the conditions under which the model applies
• Build a virtual metrology model for oxide thickness with leakage-free validation, and compare it with a simple baseline and with the physics model plus a learned correction
• Keep the furnace twin in sync with a drifting tool: estimate the drift with EWMA and Kalman estimators, update the recipe within its limits and judge the controller over the whole run sequence
• Simulate a small toolset with replicated discrete-event runs and judge a capacity or maintenance change by cycle time, throughput and utilization, together with their spread across replications
• Test a twin prototype on unseen and deliberately shifted data and complete a structured data sheet that records its intended use, validation results, limits, one data-integrity or security risk, and the action to take when confidence is low

How the course is organized

1. Start Here — Module 0: Orientation and Python Warm-up
2. Module 1 — What a Digital Twin Is
3. Module 2 — The Fab as a Data System
4. Module 3 — Physics Model: Thermal Oxidation
5. Module 4 — Virtual Metrology for Oxide Thickness
6. Module 5 — Keeping the Twin in Sync
7. Module 6 — Toolset Twin: Queues and Maintenance
8. Module 7 — Capstone: Integrate, Validate and Document
9. Course Completion

How to earn the badge

Every item in this list is required on its own. A high total cannot make up for a lab check or quiz below its minimum: for example, 6 of 10 on one lab check means that module is not complete, even with 20 of 20 on its quiz.

• Lab checks, Modules 0 to 6 (seven): at least 8 of 10 on every one (10 points each, 70 in all)
• Module quizzes, Modules 1 to 6 (six): at least 16 of 20 on every one (20 points each, 120 in all)
• Project quiz, Module 7, for the capstone track you chose: at least 16 of 20 (20 points)
• Final quiz, Course Completion: at least 16 of 20 (20 points)
• Course total: 90% or better, at least 207 of 230 points
• Post-course survey, Course Completion: submitted (not graded)

Only your best project quiz counts. In Grades the three project quizzes have their own group, "Module 7 Project Quiz (your best one counts)", which drops the two lowest scores, so the two tracks you did not take stay blank and do not count against you. That is why the course total is out of 230 points. The time logs, the pre-course survey and the optional readiness check do not count toward the badge.

This is the standard stated in the course itself, on its page "Course completion and your badge".