brownian motion

check off steps when complete.

[x] (Asher-TergeoPro-Clean) Plasma clean to remove organic debris from wafer surface. Clean four dummy wafers for dose series testing and two SOI wafers (10 micron device layer) for actual use.

[x] (HMDS-Yes-U10) Prime wafers with hexamethyldisilazane.
Notes: HMDS is a pretty cool chemical. When you’re trying to fabricate a semiconductor chip, you need to pattern photoresist onto it so you can etch the silicon in the right places. Photoresist is an organic polymer and silicon is a pure crystalline material. They don’t really like each other, so HMDS is a sort of marriage counselor with silicon arms that embed in the wafer and an organic head that is kinder on the photoresist. It’s kind of like how soap works, if soap were incredibly toxic and flammable and released ammonia upon touching water. This is why we prime the wafers in a hermetically sealed vapor deposition oven.

[x] (Spinner-Resist-L10) Spin AZ10XT photoresist onto wafers at 5000rpm for 60s to create 5 micron layer. Follow immediately with softbake at 110C for 210s.
Notes: one wafer is accidentally softbaked at 120C. this should not significantly change the result.

AZ10XT photoresist is the same consistency and color as cherry cough syrup. taste and smell may differ, but this has yet to be experimentally determined.

[x] (DirectWrite-MLA150) Expose testing pattern on dummy wafers at different doses to determine optimal settings. Use 2-micron high quality mode and a write speed of 160micron/s.
Notes: The fab is nearly empty today. I suppose that’s what happens when you work on a holiday. just me, my teammate, my mentor who has poured too much time into this undergrad project. just us, six wafers,
the humming of pipes
rows and rows
and rows and rows.
amber fields of whirring million-dollar machines.

this place looks like an engineer’s dream.
you could go insane here.

[x] (Develop-L08) Develop exposed wafers for 120s with gentle constant agitation. Rinse thoroughly with DI water and dry with N2.
Notes: the AZ435 developer solution rocks back and forth in the washing basin like waves across a suspiciously square beach. The 590nm lights make me sleepy.

[x] (Microscope-200mm-4k) Analyze dose series and determine optimal dose. Write selected dose here: 390 mJ/cm^2

[x] (DirectWrite-MLA150) Expose fabrication pattern onto SOI wafers with the previously determined dose, 2-micron high quality mode, and 160micron/s write speed.
Notes: this seems to go much more smoothly than last time.

[x] (Develop-L08) Develop SOI wafers for 120s with gentle constant agitation. Rinse thoroughly with DI water and dry with N2.

[x] (Microscope-200mm-4k) Examine wafer under microscope to ensure features are properly patterned. Measure key dimensions. Measure key dimensions. Measure key dimensions.

[x] Do not panic in the fab. Check original design. Features that are supposed to be six microns wide are only three microns wide. Features that are supposed to be three microns are only one and a half.

[x] Question what could have caused this.

[x] (Microscope-200mm-4k) Question what could have caused this. Look at the dose series in the test wafers again.

[x] Realize a small but critical error in the testing patterns. This means we chose way too strong of a dose to use in the exposure. Our mentor uncamoflages himself from the wall to appear beside us. A hunk of a man slightly creased by age, he smiles a little, lets out a breath, and tilts his head.

Yeah, it’s overexposed.
We’ll need to do this all over again.

If we tried to continue with the fabrication, our device would surely break.

[ ] Come to terms with the fact that you’ll have to do all of this again.
Notes: We’ve burned through two fabrication attempts (and several thousand dollars) already. Surely the third time should’ve worked. Did we not learn?

[ ] Come to terms with the fact that you’ll have to do all of this again.
Notes: Why, oh why, did we have to choose such a complicated final project? We could’ve been done with this fabrication two weeks ago had we chosen to make, like, a MEMS cantilever or a solar cell or something. People have written papers and spent years trying to do the thing we’re doing. What hubris it is to think that some undergrads (and, admittedly, one very battle-experienced mentor) can pull this off in a month.

[ ] Come to terms with the fact that you’ll have to do all of this again.
Notes: In fairness, this third time we tried a different wafer size and thickness and used a different resist. We had to recharacterize everything. Honestly, needing a few tries to get the process correct is probably expected.

Let’s remove the pattern with plasma. Then we can try again.

[ ] (DRIE-SPTS-Pegasus) We never get to this step.

[x] (Asher-TergeoPro-Clean) Plasma clean to remove photoresist from wafer surface.

I suppose it’s fitting that the day ends with the same machine it started with. Watch ten hours of work get destroyed by five hundred watts of electromagnetic radiation, a strangely peaceful purple glow of oxygen ions stripping away the photoresist pattern. The plasma manifests unevenly through the chamber geometry, creating a topographic map of oil-slick iridescence as it eats away at the center of the wafer and then, slowly, the edges. The resist, well, puts up a good resistance. It takes three cycles for it to fully disappear, to ash away into a stream of carbon dioxide leaving bare silicon wafers once more, pure and pristine and blank, waiting to be respun
repatterned
reborn

and walking out of the fab
stripping off the bonnet, the suit, the gloves
stepping out into the evening air, moist with rain
I feel―
[x] Come to terms with the fact that you’ll have to do all of this again.

[ ] (HMDS-Yes-U10) Prime wafers with hexamethyldisilazane.