These are my first working transistors. Specifically, they are insulated gate enhancement mode n channel field effect transistors. I also made a depletion mode FET with a conducting channel and it worked even better than the enhancement mode ones. I drew out the steps I took to make this, they’re based on Jeri Ellsworth’s work but with a few main changes. It is very important that your dielectric overlap source and drain regions on the FET, otherwise no inversion layer can be formed and the FET cannot turn on. Only a small overlap is necessary and the larger it is, the more unwanted capacitance there is. This is not normally a consideration in actual production because inherent lateral diffusion takes care of the overlap but for these large hand made devices I found you have to get very lucky with your alignment if you don’t budget extra overlap.
A huge thanks to Jeri for making her videos about home chip fabrication which got me interested in these experiments in the first place.
You can see in the center of the transistor there is a red region of silicon dioxide, this is the gate and the color indicates that it is roughly 750 angstroms thick. I would like to make the gate thinner so I can achieve lower threshold voltages, etc. but it is hard to make a truly insulating gate much thinner than that in a dirty environment because of pinholes and other impurities in the oxide layer.
A brief introduction to semiconductor fabrication processes and terminology. It is not intended to be an in depth view of any single process, but rather an overview so that provides enough information for someone to get started with making diodes and transistors at home.
Tour of my home chip fab setup in early 2017. I’ve been accumulating this equipment since October of 2016.
As of now, I am doing all of my doping with spin on diffusants. Some of which are solutions of Phosphoric/boric acid in alcohol/water that I prepared myself. I find it hard to control the dopant concentration and therefore bandgap of devices so my solar cells often have a bandgap too high or too low. I thought using an actual spin on dopant would help, so I bought some P509 phosphorus glass from Filmtronics and it works about the same as the solutions I made, except it spins on much nicer.
The colors you are seeing are contamination and nonuniformity in the thin film interference patterns caused by this thin glass layer after pre deposition of the spin on diffusants.
Blue color indicates roughly 500nm SiO2 thickness. It took about 2.5hours to grow @ 1200c in a pottery kiln style furnace with steam being pumped into it. Picture above on the left shows moisture on cooling wafer. Picture on the right shows unpolished backside of the wafer. Notice the nonuniformity in the growth thickness caused by unequal heating. This is because I don’t have a quartz boat to hold the wafers so I place them face up in the furnace. Continue reading Silicon Wafer Pics – Thermal SiO2 Growth ~500nm
First pictures show wafers with 5000A field oxide, then etched active region for the anode with HF. As you can see on the curve tracer, reverse bias conditions are surprisingly good with little leakage current down to -6 volts.
Testing of homemade solar cells/PN junction diodes. Some of them output up to .25v open circuit and a few hundred microamps at short. This varies widely because of changes in dopant concentration and wafer size. The higher dopant concentration makes a lower bandgap. Most of these devices have a band gap that is too low to be most efficient for a solar cell. Optimally I would want about 1.4ev. As dopant concentration increases, the wave functions that describe the electrons in both valence and conduction bands start to converge and overlap so the bandgap basically gets broken and the solar cells preform very poorly. I estimate about 1% efficiency on these cells, most likely less.
I wanted to try anodic oxidation because it would allow me to grow SiO2 dielectric layers much quicker and also at room temperature. The result, while using potassium nitrate solution, was a very dirty growth later that would only be suited for a field oxide on a diode and far from precise enough for a gate oxide on a FET. Possibly with more refining in the future this could save me a lot of time waiting for furnaces to heat up.