3

A few days ago I found this paper about fabricating a thin film MOSFET using ZnO at low temperature.

It seems promising to me (to do at home, some people actually made their own MOSFET at home but using high temperature furnance,) since I don't have equipment like a vacuum chamber or ion etcher machine. From what I understand of the paper (correct me if I'm wrong,) here are the steps:

  1. Dissolve ZnO powder in ammonium hydroxide and filter the insoluable things
  2. Spincoat the solution onto SiO2 wafer
  3. Place the wafer on hot plate for 30 minutes
  4. Evaporate aluminium

I don't really understand how to do step 4. I can imagine about actions in step 1 to 3, but I don't know what things are done in step 4. Can you tell me (in details if you can) what I have to do on step 4, and correct me if I'm wrong on any steps?

Steps show in the paper:

steps

Steps descriptions (section 4.1 and 4.3) in the paper:

steps descriptions

raspiduino
  • 173
  • 6
  • 1
    What's the motivation for building your own MOSFET? – Andy aka Apr 08 '23 at 16:50
  • Then I can (use / create) photoresist for lithographing larger scale semiconductor devices. Even if I cannot do that, I connect bunchs of MOSFET together to create logic devices. Even if I cannot do that, I still learn something about semiconductor manufacturing process. I know that this self made device will have a much higher cost than a commercial one, but I still want to do it myself – raspiduino Apr 08 '23 at 16:54
  • 1
    The zinc oxide was dissolved in ammonium hydroxide, not aluminum hydroxide. – JRE Apr 08 '23 at 17:27
  • @JRE Thank you! I will fix it – raspiduino Apr 08 '23 at 17:28
  • 2
    The aluminum electrodes were deposited on the ZnO surface. They probably used something like [sputtering](https://en.wikipedia.org/wiki/Sputtering), which involves plasma and a vacuum chamber. – JRE Apr 08 '23 at 17:32
  • You do realize that you have to use silicon dioxide as a substrate, right? Glass won't do. – JRE Apr 08 '23 at 18:40

1 Answers1

2

The steps are incomplete - there are precursor steps before the list even begins with step 1.

A more detailed list would look as follows:

  1. Start with a doped Si wafer - those can be bought already doped.
  2. Thermally grow 0.3μm of [silicon] oxide on it - by treating it with O2 plasma for 5 minutes.
  3. Spin-coat thin film on the wafer.
  4. Anneal on a hot plate, precipitating ZnO.
  5. Repeat steps 4 and 5 to get a sufficiently thick layer.
  6. Apply a shadow mask - it can be a PCB stencil at the large scales of the devices tested.
  7. Sputter aluminum through the mask onto ZnO.

Alternatively, you could:

  • use photoresist mask instead of a shadow mask
  • use some aqueous aluminizing solution instead of sputtering, as long as it wouldn't destroy the ZnO layer.

Steps 1 and 2 can be "skipped" by buying a pre-doped, oxide-coated wafer made to your specifications.

The biggest problem is, though that the gate is the bulk doped semiconductor of the wafer. I.e. the entire wafer is one large bulk gate. In other words: if you'd want more than one transistor made that way you'd have to either

  • dice the wafer to get a whole bunch of single transistors that you'd then either package individually using conductive bonds for gate connection, or conductively bond the dies to a hybrid substrate - perhaps even a PCB, or

  • pattern the gates, so that the entire wafer isn't just one interconnected gate

For the latter, you'd have to (semiconductor people please correct me - I'm an amateur only):

  1. Start with a Si wafer
  2. Apply photoresist pattern for the gates
  3. Ion-implant (selectively dope) the gates
  4. Strip photoresist
  5. Grow oxide
  6. Apply photoresist pattern for gate connections onto oxide
  7. Etch oxide away to expose gate connections.
  8. Strip photoresist
  9. Apply photoresist pattern for the zinc oxide film
  10. Deposit the ZnO layers
  11. Strip photoresist.
  12. Apply photoresist pattern for the metal layer - or a shadow mask.
  13. Deposit metal.
  14. Strip photoresist (if photoresist was used).

I'm not sure if the gates would need additional insulation, or would bulk undoped Si be good enough for that with adequate spacing.

At that point, going with a more standard CMOS process would be probably no harder, given the number of steps involved. If you don't care about CMOS, then NMOS or PMOS would be even easier. No dissing NMOS - some great chips where NMOS-only - all the micro revolution CPUs started that way - Intel, Zilog, etc etc.

Today's PCB solder paste stencils are usable as shadow masks for experimenting with large "homebrew" process features, so at least that's one thing you can get affordably. No matter which way you go, you'll probably want to start with something practical and well documented for "homebrew" conditions.

Probably the way Jeri Ellsworth did it would be the simplest way to go.

If you want actual chips, then just follow what Sam Zeloof did - it's a tested homebrew process, very well explained. I have little faith in replicating academic paper results, obtained in well equipped university labs, where all their mis-steps are not written nor mentioned publicly. For all we know, they may have had the hardest time getting this to work. Only the final result is what's published.

Kuba hasn't forgotten Monica
  • 32,734
  • 1
  • 38
  • 103
  • (Stupid question ahead!) You said that I can `use some aqueous aluminizing solution instead of sputtering, as long as it wouldn't destroy the ZnO layer.`. I have just searched and learn that ZnO is also conductive. So will it possible to use ZnO as the aqueous solution to create internal metal connections in the chip (more than one transistor on a single wafer)? Thanks! – raspiduino Apr 15 '23 at 13:23
  • If that is not possible, can you recommend me some `aqueous aluminizing solution` (preferably easy-to-make ones)? Thanks! – raspiduino Apr 15 '23 at 13:29
  • `the entire wafer is one large bulk gate`. From what I understand, the current will flow through the depletion region only when you apply voltage to the gate, and depletion region have a limited size. So will it possible to fabricate another MOSFET next to that MOSFET __without__ selectively dope the wafer? Thanks! – raspiduino Apr 15 '23 at 13:59