The Fab Lab Next Door: DIY Semiconductors

You think you’ve got it going on because you can wire up some eBay modules and make some LEDs blink, or because you designed your own PCB, or maybe even because you’re an RF wizard. Then you see that someone is fabricating semiconductors at home, and you realize there’s always another mountain to climb.

We were mesmerized when we first saw [Sam Zeloof]’s awesome garage-turned-semiconductor fab lab. He says he’s only been acquiring equipment since October of 2016, but in that short time he’s built quite an impressive array of gear; a spin-coating centrifuge, furnaces, tons of lab supplies and toxic chemicals, a turbomolecular vacuum pump, and a vacuum chamber that looks like something from a CERN lab.

[Sam]’s goal is to get set up for thin-film deposition so he can make integrated circuits, but with what he has on hand he’s managed to build a few diodes, some photovoltaic cells, and a couple of MOSFETs. He’s not growing silicon crystals and making his own wafers — yet — but relies on eBay to supply his wafers. The video below is a longish intro to [Sam]’s methods, and his YouTube channel has a video tour of his fab and a few videos on making specific devices.

[Sam] credits [Jeri Ellsworth]’s DIY semiconductor efforts, which we’ve covered before, as inspiration for his fab, and we’re going to be watching to see where he takes it from here. For now, though, we’d better boost the aspiration level of our future projects.

Thanks to [Byron] for the tip.

39 thoughts on “The Fab Lab Next Door: DIY Semiconductors

  1. “You think you’ve got it going on because you can wire up some eBay modules and make some LEDs blink”. Whoa, slow down, you just alienated 95% of the HaD community and 100% of all “makers”.

    1. Oh fuck off. I read these articles religiously and I couldn’t replicate most of them. My expertise is in woodworking but I frequently fix shit that others would junk. Reading an article about super easy arduino fuckery that I have no idea how to approach is still interesting. Reading the one about how to work a miter saw (a couple of weeks ago)? Not so much.
      Now, feel free to go back to that article and see if I’m there being a prick about how basic it was.
      Go on, I’ll wait.
      Point is there are a shit load of us reading this site with wildly varied interests. It is inevitable that what is elementary to one will be revolutionary to another.

    1. Recommend Floating Zone, as CZ requires discarding the expensive crucible
      “In the conventional CZ process, called a batch process, a crystal is pulled from a single crucible charge, and the quartz crucible is used only once and is then discarded. This is because the small amount of remaining silicon cracks the crucible as it cools from a high temperature during each growth run”

  2. This is really awesome, Sam Zeloof. You’re one to watch. I’m looking forward to seeing what else you accomplish soon.

    The home-brew semiconductor work so far has been minimal. J.Ellsworth’s work in semiconductor fabrication at home was excellent, but I was saddened she didn’t keep going with it. There was also the work on printed semiconductors done by Mr. Kim and Sarik but they too seem to have just stopped working on it. I hope you continue to work on it and publish your results.

    1. Thanks, I appreciate it. I find this stuff super interesting and plan to stick with it at least a little while longer considering the time and effort I’ve put into this so far. Also I have an electron microscope and possibly x ray spectrometer on the way… I will definitely publish results and continue to make youtube videos as well. I hope others start to do their own home semiconductor fab research.

  3. Nice work/gear/attitude!

    I have a couple of questions:
    1) I always thought the middle region had to be very thin (on the order of free path length / diffusion length of a charge carrier in the semiconductor), but now I try to recall why I thought this, it might have applied to only some kinds of transistor, i.e. bipolar junction? Your MOSFET gate is very wide, I thought it would form 2 opposing/back-to-back diodes in series instead of a transistor…
    2) what is the role of the quartz tube? to be smooth to slide objects in and out while withstanding high temperatures?
    3) where do you source most of your devices? eBay? or some local liquidation auction bidding site?
    4) How about a jig for scoring/breaking the wafers? I did not rewatch the video but I think you know the flat part of the wafer is in reference with the crystal orientation…
    5) For the spin coater, I assume you went with tape because of the irregular shapes of the wafer shards? perhaps coating was not uniform because the wafer is not held exactly flat during rotation? and how did that not fly off at that RPM?
    6) For the DLP projector, did you need to use a special photoresist to work with the visible light? or did you swap out the lamp? Or you were just a lot more patient with lower responsivity to visible light? or is the photoresist not that sensitive to wavelength at all?

    Please keep sharing the practical aspects of fabricating semiconductors (hopefully integrated circuits as well at some point, perhaps a ring oscillator?)

    What software will you use to design the masks? GNU Electric?

    1. 1) You’re right, the gate region on my devices so far is very large. The transistor will preform better when this region is smaller however the same field effect works regardless of the gate is 10nm wide or 1mm. I believe bipolar devices can be fabricated in a larger scale as well. They just won’t have good electrical characteristics.

      2)Yes, the quartz tube is used to provide a contamination-free environment for the diffusion to take place. Quartz is use because it is sillicon dioxide, SiO2. If normal borosillicate glass was used there would be born contamination from the glass and the semiconductor wafer would be doped with unwanted boron atoms.

      3) Some stuff on ebay. Some stuff from friends in the industry. You just have to get talking to people and tell them what you are doing and if you find the right people they will be more than willing to help out :)

      4) Good idea. I made a video on my channel about breaking the wafers along the crystal orientation but don’t have an actual rig to do it.

      5) Very possible. The mass of the wafer is so small the tape can hold it without a problem at those RPMs. Also the spin up and spin down are gradual. I am building a better spin coater with a vacuum chuck now to eliminate those issues.

      6) I used normal photoresist. I was honestly pretty surprised that this worked. You could swap out the lamp for something like a mercury vapor lamp though. You’re right, exposures do take a bit longer than using actual DUV. I haven’t played much with this or with different photo resists yet, I will soon and post my findings on YouTube.

      I will definitely keep sharing my results. A classic ring oscillator is on the list for sure. Easy way to demonstrate gain. You sound quite knowledgeable about these processes, did you work as a semiconductor process engineer or something?

      I don’t know of any good layout or design software. I’ll check out GNU Electric. I’ve been using mspaint….


      1. I only saw the first 3 videos when I posted my questions, and later saw the 10micron DLP video, where you explain mention some of the points I asked (like the UV question..).

        1) After thinking about it for a while I think what you say is true for MOSFET (which is more interesting than BJT for ICs I think), but I still think BJT (where the field effect is not used) needs the middle region to be on the order of diffusion or free path length of charge carriers: consider a wide middle region doped semiconductor, and a contact in the middle of this middle region, the middle semiconductor would behave like 2 resistors (resistivity of the doped semiconductor) with the base contact in-between, then on each side a PN junction diode connecting to collector and emitter… anyway, after seeing your 4th video illustrating the photolitography step (similar to A. Zonenberg’s paper) and seeing your pixels under the microscope you reach 10micron (congratulations btw!) this is no longer an issue anyway 10 microns is on the order of diffusion / free path length. The intel 4004 you mention in the video was at 10micron minimum feature size…
        2) Of course, that makes a lot of sense! thanks for pointing this out
        3) I guess you are better at making friends! :D
        4) about the jig, Ive been thinking about it for a while, imagine the source wafers were rectangular or square, then it would seem easy to construct a rig that breaks them in long rectangles, and then a second rig for breaking the long rectangles into squares… but since wafers are circular, it seems like a much harder task to make a cleaving jig, on the other hand, it could make some of the other steps much simpler if you had a fixed size square/rectangle of wafer, so it would be easier to mount in holders in the subsequent steps..
        5) Vacuum chuck sounds great, how do you get vacuum in the rotating chuck though? I think you also really want the fiber-optic spectrometer (I think you mentioned it pointing at a holder for the fiber-optic) ready, so you know the thickness
        6) I think yes you can in theory swap the light source, but I would hesitate a bit about mercury vapor lamp (how fast can that be switched on and off?) since the mirrors move and vibrate between color frames (resetting the DMD mirrors to neutral position, then settling them to the pixel values (until vibration stops) only when the mirrors are finally at rest is the original light source activated (if LED or …) /passed (color wheel). Also I recommend only using the blue channel and setting red and green pixel values to 0 (there may be chromatic aberration causing blurring, and the blue value is closest to UV…), unless you are prepared to hack the projectors firmware to set the binary image once without updating it for the different bits per color, but it seems like a tedious task…
        perhaps easier would be to change the light source to UV LEDs… anyway since you achieve results like it is, I’d postpone modifying the projector until you tried IC’s with the current setup… Also, I’m pretty sure you know this, but take care of your eyesight when using UV…

        “You sound quite knowledgeable about these processes, did you work as a semiconductor process engineer or something?”
        no, not at all, I’m a physics dropout… I have been obsessing about DIY semiconductor microfabrication though! so I read a lot…

        Do you have some kind of roadmap, are you managing priorities within this project?

        If you need idea’s or help (anything, say configuring the settings in Electric for your custom technology node), I can’t guarantee my advice will be good, but I’m willing to try and help out, everything you do (both succeeding or failing) would be valuable lessons for those of us who wish to follow…
        I’ll send you an email soon, so you can contact me on this subject whenever you wish…

        Oh, although your focusing seems fine already (given the pixelation in the microscopy image) I was thinking about a kind of thin plate with a rectangular window, so the wafer gets pressed up against the window but is stopped by its border (so you lose a little area on the wafer), then instead of a wafer, you take a camera sensor, or a webcam with lens removed, and push it up against the window as you would with the wafer, then you can focus the projector image, and see the resulting focus live while changing focus distance… given the strong heat of the projector (which may have been on for a while, or just turned on), it might be wise to make the projector-macro-lens-window holder from a material with low thermal coefficient of expansion… then you no longer have to focus in the future, just press the wafer up against the window… You might also want to surround the wafer being lit with black(ened) material (or mirrors pointing to blackened material) to avoid diffuse reflections of the intense projector image to disturb the intended pattern…

        I really look forward to seeing you succeed

      2. We tried using a ferro fluid to make the vacuum seals in a very high vacuum. There were 4 or 6 of them. I can’t remember what the fluid was but it didn’t contaminate our oled coatings

  4. Only 12 minutes into the first video but so far so good Sam. Perhaps it is mentioned at some point but this looks a little like a university, or resume building type endeavor which I really like. I will finish watching the rest later today.

    Nice work!

  5. Fab lab FabLab?
    Sam’s setup is for making silicon stuff, a fablab is a ‘maker space’ with a specific set of CNC tools (3D printer, lasercutter, vinyl cutter and CNC router), a concept thought up by Neil Gershenfelt of MITs CBA.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.