Week 14 — Wildcard Week - Microfabrication

Image Wafer: Microfabrication Inspired by MIT.nano

I walk past the MIT.nano building very often, and one of the things that always catches my eye is this One.MIT project: a permanent display tucked into a quiet corner of MIT.nano. At the time of this writing, there are two disks on display. On the newer and larger disk, nearly 340,000 names of MIT alumni, faculty, and students have been densely etched onto a surface no larger than a DVD.

In addition, there’s a similar disk on display at the MIT Museum, paired with an interactive setup that helps visitors find their names if they’re part of the MIT community—which is pretty cool.

One interesting detail is that the cutoff date for the names is 2024, before I attended MIT. Yet I was still able to find my name (on the disk displayed at the Museum, that is—there’s no way to visually distinguish individual letters at the nanoscale). I imagine this means that identical names are only etched once, and as it happens, I do know someone with my exact same name who graduated from MIT two years ago 😀

Inspired by that, I decided that among all the interesting options for this assignment, I would choose microfabrication and try to “draw” an image on a disk in a similar way.

Understanding Silicon Wafers

Wafer: A thin, flat, ultra-pure substrate mostly made of silicon(~95% of the semiconductor industry) used as the base for microfabrication and integrated circuits.

Why silicon?
• Semiconductor whose conductivity is tunable via doping
• Forms high-quality native oxide (SiO₂) for insulation
• Abundant and relatively low-cost
• Mechanically strong and thermally stable
• Supported by a mature industrial ecosystem

Dimension and Purity
• Thickness: ~500–800 µm (≈ credit card thickness)
• Diameter: 4", 6", 8", 12" (300 mm standard)
• Electronic-grade purity: 9N: 99.9999999% 11N: 99.999999999%
• Clean rooms are essential: modern fabs operate at Class 1 or better (≤1 particle per cubic foot of air).

Why wafers are disk-shaped
• Cylindrical crystal growth → circular slices
• Rotation-based processes require symmetry:
  • Spin-coating photoresist
  • Uniform heating
  • Centrifugal drying
  • Symmetric chemical flow
• Disks coat evenly; squares cause edge buildup and stress
• Industry lock-in: $10–20B fabs optimized entirely for round wafers

Unused or defective regions
• Edge regions, damaged dies, and failed chips are scrapped
• Some material is recycled as raw silicon feedstock
• Crystal perfection cannot be restored → never reused as wafers

Standard Microfabrication Flow（sourced gathered from ChatGPT and other online source and reorganized here)

1. Start with a wafer: Clean, inspect, and prepare a single-crystal silicon wafer.
2. Form base films: Grow or deposit thin layers (e.g., SiO₂, Si₃N₄, polysilicon, metals).
3. Coat with photoresist: Spin-coat liquid resist and soft-bake to form a uniform film.
4. Pattern by lithography: Align the mask, expose with UV light, and develop the resist to create a pattern.
5. Etch or modify material: Use wet or dry etching to transfer the pattern into the underlying film, then strip the resist.
6. Doping: Introduce dopants by ion implantation and activate them with thermal annealing.
7. Repeat for multiple layers: Alternate film deposition, lithography, etch, and doping to build transistors and interconnects.
8. Planarize: Use chemical mechanical polishing (CMP) to flatten surfaces for further processing.
9. Passivation: Deposit protective layers and open contact windows over bond pads.
10. Wafer test: Electrically test each die on the wafer and map good versus bad chips.
11. Dicing: Saw or laser-cut the wafer along scribe lines to separate individual chips.
12. Packaging and final test: Attach dies to packages, connect them (wire bond or flip-chip), encapsulate, and perform final electrical tests.

I’m using the open-source imagewafer project to convert a bitmap image into a GDSII layout using Python and PHIDL. The link to the site is here: imagewafer GitHub repository .

• Goal: Take an image (like mitnano.jpg) and convert it into a GDSII layout made of tiny “pixels” (rectangles or shapes) that you can write on a wafer using tools like the MLA150.
• Toolchain:
  • phidl – a Python library for creating GDS layouts (often used with KLayout, lithography, photonics, etc.).
  • imageWafer.py – a ready-made Python script that:
    • Loads the image.
    • Converts it into a grid of pixels.
    • Turns each pixel into geometry in a GDS file.
    • Optionally adds some text.
  • Output: imageWafer.gds (or similar) → can be sent to the MLA150 or viewed in KLayout.

I did have to install the phidl library first. I used pip to install it in my Python environment:

When I first run the code there were error reporting, after some investigation, it appears to be the issue of a font library. I made changes in the code to change the font setting from "Calibri" to "DEPLOF", which is PHIDL’s default font, and the export succeeded!

This is the code I used to make the final conversion. I modified the code I downloaded from the open-source imagewafer project. Changed a few parameters, including the default font and the file path used to load photo/pictures.

Unfortunately I don't have lab access, so I sent the converted file to Aijia and asked her to do it for me. The file was big, around 170MB!

Since a wafer is made of silicon and is, at its core, refined sand, looking at this immediately reminded me of William Blake’s Auguries of Innocence.