Follow the project

On this page

Overview: purpose, approach, and current milestones

The goal of Gotham Silicon is to make fabrication of simple MOS electronics easy, fast, and cheap.

A gloved hand holding a small patterned silicon chip
View full-size photo

Process development is happening at two university cleanrooms in the US: Columbia Nano Initiative, and Berkeley’s Marvell Nanofabrication Laboratory.

Recordings of lab sessions

Recordings of lab sessions are available on my YouTube channel.

Tools with documentation

Furnace

A ~$100 tube furnace that reaches over 1,000 °C and can be assembled in two hours if you know what you’re doing.

Ingredients

  1. MMasters ceramic fiber insulation blanket (24 x 8 x 1 in) (Amazon)
  2. Master Wire Supply Kanthal A1 wire (22 gauge, 50 ft) (Amazon)
  3. beinkmake insulating ceramic support rods (7.87 x 0.08 in, 10-pack) (Amazon)
  4. CDMALL AC motor speed controller (120 V, 15 A) (Amazon)
  5. GTJ silicone baking mat roll (12 in x 6 ft) (Amazon)
  6. Quartz glass tube (1 in OD, 0.866 in ID, 12 in long) (McMaster-Carr)
  7. Ceramic washers (1.062 in ID, 2.5 in OD) — 2 required (McMaster-Carr)
  8. K-type kiln thermocouple probe (0–1300 °C, 3.3 ft extension wire) (Amazon)
  9. HBN plug-in watt meter with backlight (Amazon)
  10. Proster dual-channel thermocouple thermometer with 2 probes (Amazon) or SafeSense TempTouch Wi-Fi thermocouple sensor (Amazon)
  11. High-temperature polyimide tape (1 in x 33 m) (Amazon)
  12. GUBCUB two-way ceramic terminal blocks (25 A, 380 V, 5-pack) (Amazon)

Steps

V = IR

W = IV

360 W = 120 V × 3 A

120 V = 3 A × 40 Ω

  1. Measure and cut the wire

    Measure and cut a ~40-ohm length of Kanthal A1 wire. Remember to err shorter rather than longer, because you can compensate with the AC motor speed controller. If you cut too long, such that you can’t get enough power using the 120 V wall outlet, then your furnace might not get hot enough. Shorter is also easier for winding. If you cut way too short you will not have good control over the system.

  2. Wind the heating element

    The next steps require some finesse. Find something approximately the thickness of a pencil, but ideally round, more rigid, and slightly longer—something like a threaded rod or dowel.

    Uniformly wind the wire around that rod until it is the desired length for your heating element (~6 inches if you are using the 8-inch-wide insulation listed above). Once you’ve wound 6 inches, slide the winding off the rod; it should hold its shape.

    Start winding another 6 inches with ~2 cm of straight wire between the windings, until you’ve consumed enough of the wire. Make sure to leave ~6 inches of unwound wire on either end to have ample room for electrical connections.

  3. Secure the coils to the insulation

    Place the ceramic fiber insulation flat onto a working surface. Use gloves and be outdoors or in a safely ventilated setting.

    Insert one of the ~6-inch wire leads through the insulation 1 cm in from a corner, pulling the lead through until the first wound coil stops it from moving further. Place one of the ceramic rods through the middle of this first coil so it can’t bend. The rods are there only to provide structural support; the Kanthal A1 wire is the heating element.

    Make a switchback pattern with the subsequent coils, ensuring that in total the switchbacks don’t take up more than 7.8 inches of the ceramic fiber insulation (2.5 inches, the outer diameter of the washer, × 3.14).

    Insert ceramic rods into each coil, and cut ~2 cm lengths from your original spool of wire to use as staples to secure the coils into the insulation. Be careful to ensure coils and staples aren’t at risk of shorting to each other.

    Once all the coils are safely secured, thread the ending lead through the ceramic fiber insulation and pull it taut, like you did at the beginning.

    An earlier furnace heating element secured in a switchback pattern on ceramic fiber insulation
    This image is from an older furnace iteration. It does not have the described coils with rods in the center, but it does show the switchbacks secured by metal staples. Hopefully it gives you a sense of how things might look. Sorry for not having an up-to-date photo of this step. View full-size photo

    A diagram is coming soon to better explain this.

  4. Position the quartz tube and washers

    Once the coils are secured, place the ceramic washers around the quartz tube. Place the tube centered on the insulation, protruding 2 inches out from it on either side if you bought the insulation above, with each washer on either side of the coils, ~1 cm from the end of the insulation.

  5. Roll the insulation

    Roll the insulation around the washer/tube system until the insulation totally surrounds it, then continue rolling until you run out of insulation. Keep threading the leads through the insulation so they have a path out from the center.

    Be careful as you roll that the coils stay put. If they get dislodged and close to each other, they may continue to move around and immediately or eventually cause a short. If you notice the coils are too close, unroll the insulation and adjust them, securing with additional staples, changing the coil spacing, or whatever else is necessary to ensure this risk is sufficiently mitigated.

  6. Wrap the furnace

    Wrap the whole insulated cylinder with the silicone sheet, cutting it to size, and use Kapton tape to prevent unraveling and keep everything together. The leads should puncture the silicone cover so you can connect to them.

  7. Connect the electrical leads

    Connect the heating element’s electrical leads and the hot and neutral conductors of a modified electrical outlet cable (any old extension cord or power supply cable) using the ceramic electrical terminal.

    The assembled tube furnace outdoors beside a temperature meter
    View full-size photo
  8. Add end caps

    Add silicone or Home Depot tile end caps on either side of the tube for fanciness.

    Furnace end caps · 3-second video. View on Google Photos
  9. Connect power and monitor the temperature

    Plug into your AC motor speed controller, power meter, and a wall outlet.

    Slowly ramp while monitoring the temperature in the tube.

If you built the furnace right, you should be able to swap out tubes fairly easily because they are only touching the ceramic washers. You can be fancy and have a tube for everything: a temperature calibration with the thermocouple in its tube, a dry gate oxidation tube, a P-type doping tube, an N-type doping tube, a wet oxide tube, an anneal tube, etc.

As you make devices, remember that cleanliness is important. During high-temperature steps in particular! There are so many ways a home or cheap chip fab can go wrong. Sensitivity to contaminants during thermal processing is one to pay attention to.

A potentially much easier thing to do than build your own furnace is to just buy the quartz tube from McMaster-Carr and buy a couple of canisters of MAPP welding gas, slowly heating the tube or a region of it to the desired temperature. Be careful not to make it explode from significant temporal or spatial temperature gradients. I haven’t tried this, but I think it would get you hot enough to do some doping.

Two gas torches positioned around a quartz tube
View full-size photo

POV recorder

A system for POV recording/streaming, nominally for lab work.

The benefits of this system over smart glasses are that the battery lasts as long as you want, you can script whatever functionality you need, and you can change the sensor/lens to your liking.

Ingredients

  1. Walfront OV5693 USB camera module (5 MP, autofocus, 180°) (Amazon)
  2. USA Gear camera chest harness with quick-release buckles (Amazon)
  3. Nitecore NB10000 Gen 4 power bank (10,000 mAh) (Amazon)
  4. Hagibis USB-C splitter (TUA08 USB + PD, 100 W, 10 Gbps) (Amazon)
  5. Samsung phone I used (inactive listing) (eBay)
  6. Tape
  7. Your Android smartphone to use as a control surface
Back to top ↑

Process research agenda discussion

It is possible to make simple custom devices very quickly by designing your process specifically for that purpose. Some ways I approach this:

  • Stock pre-implanted wafers with oxide and polysilicon layers grown/deposited in advance to avoid any long furnace cycles at fabrication time.
  • Use spin-on dopant with rapid thermal anneal furnaces instead of sending wafers to external services for implant whenever possible.
  • Use SOI wafers and mesa isolation instead of more complicated and time-intensive STI or LOCOS techniques. SOI also gives you the ability to do work function engineering that could allow for junctionless devices, further minimizing the need for implant/doping steps because you don’t need to form wells in oppositely doped bulk silicon.
  • Instead of exclusively relying on maskless lithography, which inherently means your exposures have to be done in sequence, you can instead rely on a combination of maskless and masked lithography. As soon as you receive an order, you can start fabrication immediately by doing the first layer on a maskless setup while initiating mask writes for all subsequent layers in parallel across multiple mask writers, so all subsequent lithography can be done more quickly.
  • Obviously, but importantly, minimizing process steps gives three wins: shorter-loop process development, faster fabrication, and improved fundamentals of yield (number of steps × reliability of steps). It is not a perfect generalization, but the shortest process that does X is a good approximation of what I am trying to do.

The following are representative run sheets for the processes I am researching and aiming to commercialize.

Each runsheet includes an interactive step explorer, diagrams, and a printable cleanroom sequence. Runsheets open in a new tab.

Berkeley NMOS test processes

Testing Poly gate NMOS fast

Supplied oxide / poly stack · first transistor screen after one lithography level

Testing aluminum gate NMOS fast

Two lithography levels · phosphorus SOD · evaporated aluminum gate and contacts

Berkeley NMOS production

Four patterned levels · self-aligned polysilicon gate · phosphorus spin-on doping · one interconnect metal

Berkeley CMOS production

Six device-pattern levels · two spin-on dopants · self-aligned source/drain · one interconnect metal

Back to top ↑

Some good moments from the past two years of chipping!

Pepsi and aluminum contacts

Pepsi (and many other sodas) have a small amount of phosphoric acid. Enough to etch away the native oxide that grows on your aluminum contacts in ambient conditions, so dipping your chip in the Pepsi freshens it up for probing…

A hand holding a small cup containing a patterned sample beside a Pepsi bottle and bench equipment.
View larger photo · Google Photos
Back to top ↑