Printed Semiconductors for 3D Metal Printers

MetalicaRap will soon be the world's first self-printing 3D metal printer. My project will one day let it print its own integrated circuits.

Similar projects worth following
A lot of people are excited about self-replicating 3D printers, but no one has had the guts to ask, how can a printer ever print its own integrated circuits (microchips)? The MetalicaRap project is aiming to build a self-replicating 3D printer that uses electron beam additive manufacturing to make its own metal parts, and solar cells. I realized that MetalicaRap's vacuum chamber can be equipped to print integrated circuits using ion implantation and chemical vapor deposition. Commercially, these processes use a cheap but complicated process involving a photoresist mask. My design uses a steered ion beam to do away with the mask which is far more expensive on a small scale. This allows 3D printers to print their own silicon microchips onto a blank die. It also allows anyone to make their own custom microchip. I started 3 years ago by replicating Ellsworth's DIY silicon transistors, and now I am going to build an ion implantation instrument using parts from a mass spectrometer.

2016 Hackaday Prize - Design your concept

This project is a single step toward a huge goal - a self-replicating machine. Others have already taken on the challenge of self-replicating the mechanical hardware of a 3D printer, I will take one small step at a time toward replicating the electronics.

How can a 3D Printer Replicate its own Electronics?

If MetalicaRap does good on its promise to build an electron beam additive manufacturing 3D printer, that printer will have a high vacuum in its main build volume. It will also have the correct optics to guide a beam of electrons onto a tiny spot. If it can do that, it can also guide a beam of ions onto a small spot. By placing a blank silicon die in the build area of MetalicaRap, the focused ion beam can selectively n- or p-dope silicon and create the fundamental building blocks of integrated circuits - transistors, diodes, capacitors, resistors, and more. Metallization layers can similarly be deposited in the vacuum chamber.

Then why does a semiconductor fab cost so much money?

Because it uses photolithography to manufacture a lot of chips fast and simultaneously. Photolithography is an extremely delicate and high-initial cost process. Modern semiconductor fabs also try to make transistors on the smallest possible scales to pack as much into a chip as possible, which greatly increases costs and engineering requirements.

Can you really make semiconductors at home?

Jeri Ellsworth already did it (despite everyone telling her it was impossible), and I replicated her results a few years ago. She made her masks by hand and used diffusion doping, which means that her transistors were too big and inconsistent to make an integrated circuit. With ion implantation, there problems are reduced.

So what exactly are you making?

The specific goal of this project is to demonstrate ion implantation using standard scientific equipment, and then going backward and seeing if there is any way to simply or improvise some of the equipment so that it can be printed using electron beam additive manufacturing, and so that the process can take place at higher pressures. In the future, my design can be incorporated into MetalicaRap so that it can replicate its own circuits.

Another challenge that I will explore is how much can 3D printer microcircuitry be simplified so that it can be printed at larger (and therefore more reliable) scales.

Why did you wait so long to do this?

What enabled this project to go forward was acquiring a QSTAR XL tandem mass spectrometer, which has many of the parts needed for an ion implantation design.

Testing Vacuum Pump Motor.xlt

Test results of (1) the resistance between the 8 wires coming out of the motor windings and (2) the result of various diagnostic tests preformed on the motor

xlt - 13.50 kB - 04/25/2016 at 04:09


  • 10 × N-doped silicon wafers From
  • 2 × Dilute hydrofluoric acid Wink Rust and stain remover
  • 2 × Rotary vane fridge compressors
  • 1 × Mason Jar For physical vapor deposition
  • 1 × Copper tubing For physical vapor deposition

View all 19 components

  • Current Progress

    Misha Dubrovsky04/25/2016 at 04:50 0 comments

    I have already repeated Jeri Ellsworth DIY NMOS transistor experiment. A lot of the pictures and components listed in this project are from that experiment.

    The next step is getting a working vacuum system. The relevant hardware I was able to salvage from the QSTAR XL tandem mass spectrometer is (1) a Varian DS 602 2-stage rotary vane pump (2) three Leybold turbomolecular pumps (3) two vacuum pressure gauges and their controller (4) anode, cathode, grind, etc. integrated power supply.

    Currently, I'm trying to get the Varian DS 602 vacuum pump working. When I received it, it was running very hot and blowing a 20A breaker after a few seconds. I disassembled the actual pump portion and found nothing wrong at all. Then I tested the motor. There are no visible issues with the motor but it runs unreliably. It is a somewhat unusual and complicated design (dual-voltage capacitor-start capacitor-run squirrel cage induction motor with starter relay) and I have not been able to find the schematics. I drew out the schematic of the external wiring and recorded the resistances between the coil wires. From there, I inferred the connections inside the windings. Now I have disassembled the motor several times and tried various things (detailed in the spreadsheet attached) to identify the problem winding, but with little luck so far. I am also shopping around for potential replacement motors at surplus stores.

    Once the vacuum pump is back online, the next step will be to check the turbomolecular pumps. I have already contacted some companies for help with that step, and I am in the process of designing an Arduino sketch to control the pump frequency converters.

    Then I will build the main vacuum chamber with electrical feedthroughs, view ports, venting inlets, and vacuum gauges. After all of that works, I am ready to build the ion implanter!

View project log

Enjoy this project?



Jerry Biehler wrote 11/20/2016 at 04:49 point

Looks like a Leybold TW700 series pump? The turbo pump does not need an arduino to run it. Most drives have a standard switch input for on/off. Just supply power and close the switch. You can spin the pump up at atmosphere to verify it starts, you only need to run it a few seconds to verify. Newer pumps will time out when the controller sees the pump cant reach a certain speed after a certain amount of time. 

Usually old turbos out of mass specs will run, they never see a dirty environment like a lot of pumps that come out of semiconductor manufacturing equipment. Those can have some really nasty residue in them.

Be very careful with turbos. There is a huge amount of kinetic energy in a pump this size when at speed. If you open the chamber to atmosphere while running very bad things will happen. Even a short burst of pressure from something like air trapped between two valves can be enough to do damage. 

  Are you sure? yes | no

Sci wrote 11/16/2016 at 06:12 point

Hey, just a tip! If the rotary pump needs new parts, you'll probably be better off getting a refrigeration service pump. For the price the science-grade vacuum suppliers charge for spares, you can get a brand new HVAC servicing pump of equal or better spec in terms of pumping speed and ultimate pressure. It's only a backing pump and they're pretty interchangable.

  Are you sure? yes | no

Jerry Biehler wrote 11/20/2016 at 04:42 point

Oh gosh no. Do not get a refrigeration pump, these things have very low throughput. 

The motor looks to be a standard C mount motor which can be easily replaced. You can probably find a used motor for less than $50 on craigslist or ebay.

  Are you sure? yes | no

Sci wrote 11/20/2016 at 06:14 point

HVAC servicing pump, not a refrigeration compressor pump. As you imply, compressors aren't useful for very much more than tinkering with and don't have much throughput. I got a large HVAC servicing pump (rotary vane) to back a diffusion pump system, and it's spec exceeded what was required. £130 new, iirc. A roughing pump of similar spec from a scientific supplier was priced 12 times as much.

  Are you sure? yes | no

Jerry Biehler wrote 11/20/2016 at 06:21 point

I was not referring to a compressor, but a HVAC service pump. The are extremely limited in capacity due to the flare ports to connect to manifold gauges. He has a pretty decent sized turbo there, you dont want to back this with a fridge pump, if there is any significant gas load in the chamber the foreline pressure will spike which could kill the pump. You can get away with a lot with diff pumps. Worst case scenario with a diff pump, assuming you are not running hydrocarbon pump fluid, is you get a mess. Worst case with a turbo, the thing shreds itself internally and breaks itself off the system ripping lines.

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates