I want to make chips, period. In order to do that, I have to build specialized tools to be able to do so.
1.- Vacuum chamber for DC sputtering deposition. (completed) 2.- HVAC manipulator with glass feedthrough (completed). 3.- 1200ºC tubular oven with integrated timer. (completed) 4.- Spin coater (completed, missing proper buttons) 5.- Hot plate. (completed, requires housing) 6.- Laser milling of PMMA in Silicon for first tests and 100µm feature size (completed) 6a.- Micro focus adjuster (completed). 6b.- Adding a camera for fiducial location (completed). 7.- Fume hood (completed)
Special Thanks to Niklas Fauth (https://hackaday.io/niklasf) for his invaluable help sourcing future testing hardware and other tools for future research.
To build semiconductors at home, you need a 1000ºC oven with controlled atmosphere, a way to diposit metal on top of silicon and a safe way to handle deadly acids. There are also complementary tools, like a spin coater or a hot plate, needed to apply acid resistant film on top of the silicon, to etch away the patterns.
All that can be made at home with some ingenuity.
Two pdf "Books" have been created from all the research done for this project. (stored in the "Files" section.
Semiconductors @ Home - Compendium! (all knowledge on the tools required for each process, the main focus of this project)
Semiconductors @ Home - Cookbook! (a step by step guide to use the tools fabricated, under constant update)
Below, some image highlights of the different tools and processes:
First diode, using Boron acid dopant in N-type silicon (phosporous):
For smaller resolutions, the plan is to investigate UV-curing resins and DiY e-beam maskless litography on PMMA, repurposing a CRT tube with a custom controller (already in prototype stage)
Once the semiconductor process is refined, vacuum wirebonding and DiY chip carriers will be tested.
First noticeable diode effect in N-type (phosphor) wafer, using Boron.
Right now I'm stuck with N-type (Sulphur) doped wafers, so I'm using boric acid as boron source. When cooking it, it produces borosilicate glass on top of the wafer, wich, if you use too much of the dopant, makes for such a thick film that can't be really etched away. Luckily, a small window opened in the center of the test, and I could access the boron doped part. ^^ (after diffusion, before etching)
When we left, this was the general idea I wanted to achieve:
A magnetic HVAC interface using a (modified) glass test tube for vacuum holding.
First thing I adressed was the driver holder/actuator. Gear/rack mechanical coupling proved impractical, so I switched to bowden drive:
I had thin braided steel cable in my workshop, for pneumatically pulling things in other projects. What I did not have was bowden tube of any kind!
Coaxial cable, help me!
UHF cable core is usually made either of poliethilene foam or, in higher grades, from PTFE. I wasn't as lucky as the second choice, but for short distances and small forces, PU foam would work fine. (for future projects I did order PTFE tubes)
A linear rail was added to the back to attach the motion driver to the test tube and provide Z axis movement/control.
Wich would also be bowden controlled:
Springs on the other side of the bowdens would provide cable tension. However, altough the bearings could take it, they exerted much lateral force in the pulleys, binding the rotary motion, so, a second version with zero radial net force was implemented.
The axis of the push/pull force in the cable sit in the same plane, nullifying the force exerted on the pulley (technically, it generates a twisting motion as the cable does not alignt with itself, but the bearings can cope with that easily)
Also, the fixed test tube holder didn't really work, as each test tube is slightly different, so, the linear guide holder was made as separate pieces, making the tube supports adjustable.
Teflon ball cages where scissor cut so the balls would sit properly in the pulleys:
Small sections of brass tube would be flat clamped to the cable to make connections.
After that, it was time to make a hand controller for it. So I mostly went nuts with the chamfer tool in fusion 360:
That hand piece would get attached to a linear rail with an adaptor, wich in turn would be directly screwed to the sputtering table.
The whole hand will move the Z axis, whereas each finger will have direct control of one of the pulleys with these finger cable pullers:
And how does it work?
AMAZING, if I can say so:
Here's a bottom view of the assembly coupled to the vacuum chamber:
All the bowdens where cut to the minimal lenght and lubed with silicone oil. For upper guidance/parallelism a PTFE guide for the metal parts was cnc cut:
So what does all this accomplish?
I now have a modular system with 3 coaxial axles with Z movement that I can employ for anything inside the vacuum chamber, WHILE it is operating.
I can have different motion groups for different purposes and all I need to do is remove the internal magnet assembly, wich slides out into the chamber, and put a new one, with different end effectors/holders/targets.
Here, for example. I'm spot welding a small strip of stainless, in provision for the small arm that can move samples around:
Instead of an arm, I can use it to put a plate over the samples and cover them until the sputtering is stable and clean.
I can have one arm with targets, and switch them at my convenience, while a second arm covers the samples meanwhile the changes (leftover sputter of a different material) stabilize.
There are infinite possibilities for this, It is just a matter of what work I need doing inside the vacuum environment.
The glass test tube is hand-modified with an Oxy-torch in the lathe, so you can do everything at home, no need to depend on anyone for weird pieces.
Until now, I had been doing the laser litography tests with the machine as is. Nailing the focal point was tricky, if I ever found it.
Once I started having better results with the technique (right in the above image) it was clear I needed a better way to find and mantain focus with the silicon.
The holder also has to be hollow, since the laser mostly goes through the Silicon, and having a surface just below the wafer, could prove problematic.
So, holes where drilled...
And a vertical linear rail was added:
The design requirements dictated the shape, edges and lengths, so there wasn't much to the design:
This piece, working near a laser, should not be made out of plastic, however, for quick tests, I 3D printed it to have a feel of how it sat in the machine. (Just ignore the bearing, it was meant for something else, but it was not needed in the end.)
I will be using the cutting height adjusters, however, just as they where, it could probe flimsy for precision height adjustment, so the screw was preloaded with a spring and an axial bearing to prevent twist resistance:
Everything looked great, so I embarked in the machning. Nothing in it was size critical, only the mounting holes, so I preferred to hand machine it, as my mill can remove much more material than my cheap CNC router can.
Holding is done through flat springs, sitting against the small 1mm ledges in the piece, and made pretty much like the cover in the hot plate:
(The ones in the image are just temporary, better ones will be made. To prevent chipping, the side in contact with the silicon is rolled, so it presents a smooth surface.)
Up until now, I had been using the oven at a lower temperature, as a makeshift hot plate/convection oven to dry thin fims.
However, as I begin to use the oven for it's intended purpose (growin SiO2 and difussion) it becomes very counterproductive to use it for other things. So, a hot plate to dry the thin films was devised.
A suitable chunk of aluminium was procured from the workshop, wich offered enough space for multiple pieces and could hold in itself the heating cartridge I had around.
Drilled and milled:
For the K probe, the retention screw had a weird thread I didn't had a matching tap for, so I ended threading the probe itself to M5 and screwing that into the aluminium block. Having it's head sitting just 3mm below the surface of the plate. Plenty of thermal compound was employed.
The heating cartridge was retained with a screw and plenty of thermal compound too.
Finally, I wanted to add some termic isolation to the bottom, so the assembly could be made compact. Ceramic matt tends to be fragile, so mechanical subjection is not recomendable. Instead, I scissor cut a piece of solder paste stencil and marked it with a cutter.
Clamping it into a vice, it was first hand bent and then shaped with a nylon mallet:
The inside corners where bent using a spacer:
With that and careful measurement, a super nice bracket for the ceramic matt was done:
To further isolate the electronics from the plate, sheet metal legs where spot welded to the plate:
With that, but pending a different temperature controller with SSR capabilities, I connected it to an old controller I had around, and for now, I have a sketchy, but working hot plate!
Once I get the definitive controller, everything will be made much compact, with the plate on top of the controller + SSR, a nice case and some form of heat shield so you can't accidentally touch the hotplate sides.
With mostly everything ready to begin tests, bits of safety remained to be solved.
Since I will be working with dangerous chemicals, I didn't want to have the small bottles of them hanging around all over the place.
Thus, I came up with this:
This holds onto the bottle neck in the front and has a suction cup in the back, so it can be attached to the wall of the fume hood. It is also interconnectable with other pieces, so it can form a neatly arranged array:
They are easily and fast 3D printed with a 0,6mm nozzle:
They should be improved by changing the 2D side connector (isc clamp) to a 3D version (ball clamp) so they can't slide out, altough that would be difficult with bottles in each holder.
Huge array, in preparation of semiconductor tests:
With the excellent result achieved with the improvised spin coater, it becamse sacrilegous to leave it at that, dangling wires and crappy mounting.
Thus, I set myself on the task of building a nice enclosure. First I quickly modelled everything in Fusion:
A thick base was added, so it's weight could act as vibration dampening. It would have a recess for the vacuum pump, as it was 8mm thicker than the power supply, and also a hole for the vacuum adapter for the motor to fit in. Resuming, I wanted the thinnest possible unit.
With that, I set up workshop to machine, cut and weld everything.
First, pump recess machining:
The acuum adapter connector was changed to a slimmer one, and thus, the shape of it's recess was also changed on the fly:
Pump clearance with the power supply:
I also added suction cups.
With that done, a stainless case began to take shape:
Fully welded and blended:
At this point, since I didn't had neither AC connectors nor panel mount buttons, I decided to leave the case like that, and just attach the controller to the outside with double sided tape (over kapton for easy removal) and wait the components.
I finally got into building the spin coater for the project, however, even I was surprised about the way things went.
All I knew is that I wanted a small spin coater, something that could do chip sized things and maybe something a little bigger, but nothing more.
I started with a powerful 40*40mm brushless fan from Delta I had around. It has ball bearings for smooth operation and also it's directly PWM controllable and has smooth start built in. (more about that later)
The fins where removed to reduce current consumption:
With that, I picked up a simple lcd pwm controller from ebay and hooked it up:
It worked well enough, so I faced the rotor in the lathe.
Using double sided tape, I tested a mix of PMMA and acetone:
It worked well enough, so I quickly cut an acrylic panel and bolted everything together:
This is a bad moment to discover that you have absolutely ZERO panel pushbuttons and you have to leave the start button dangling underneath:
Quick test with double sided tape:
After 25 minutes in the oven @ 150ºC it looked nice. Since I'm using acetone as solvent, it evaporates too quickly, leaving the white marks. With proper technique an better solvents, it should work fine.
Given the results, I was going to leave it at that, but the project had other ideas.
Make it better because your workshop says so.
I happened to have some capillary tubes I ordered for the manipulator, but two of them where iron, not stainless. It also happened that one of them had the right diameter (3mm) totally interchangeable with the motor axle!
I just needed a vacuum pump, but it was going to be either very expensive, or take waaay too long to get here from china.
I also happen to NOT know very well what my "junk" drawers hold. This time, the pneumatics section did had a 24V vacuum pump ideal for the project!
So, I pretty much was good to go and try the vacuum chuck thingy.
The axle was changed by simply hammering it in, no need to remachine anything.
...and extended. I left it longer than the original to try to protect the ball bearing from any coating that could seep in.
A vacuum tube adapter was machined from scrap aluminium:
I used double sided tape to hold it in place and test, using an o-ring and some cardboard as spacers so it could not vibrate loose:
And some grooves where machined to hold two standard 10 and 16mm OD o-rings:
But...does it work?
Hell yeah it does! (5550 rpm)
It also works well with odd shaped glass:
But does the vacuum chuck make any difference, or is it working by sheer chance? Let's choke the vacuum line and see:
As for power, I had lying around a 24V 1A switching power supply:
Initially I was also looking for a 12V power supply for the spin motor, but a quick measurement of the consumptions later:
(choked vacuum pump)
(spin motor with chuck ON)
With those kinds of powers, I can just simply throw a 7812, heatsinked for the lulz, and call it a day. Also remember that the smooth start prevents the spin motor from drawing lots of current, killing two birds with one stone.
As final note, let's talk about speeds.
This motor is not a normal computer fan. This is a high power, 18500 rpm nominal, 800mA brushless motor. Usually, the final RPM's @ 100% power depend on the density of the air and many other variables, so using it without modifications would require direct measurement and/or characterization.