Doping: The Most Important Part of Making Semiconductors 

yeah the first time i heard it i was like "ooooooooooh i get it now"

same analogy written in Solid states of physics by Donald Neeman.

I appreciate it :)

Thank you for the compliment :) I did my best to make sure I didn't go into the weeds too much

Haha if you do please show off your work!

Thank you! I am trying my best

There are actually some startups with the goal to make small scale fabs that fit in a single room. I am hopeful they succeed

@@projectsinflight Are you refering to semiconductor manufacturing startups or the type that provides with development tools for electrical engineers and outsource manufacturing to an actual fab? I'm talking about Tiny Tapeout here for example. If you mean the actual manufacturing startups would you mind sharing some info here please?

Just wait... This will be a core part of von neumann replicators!!!

@@pavlokachor6544lookup Atomic Semi by Sam Zeloof 😉

@@pavlokachor6544 checkout “atomic semi”, they’re trying to do what 3d printing is to injection moulding but for semis.

Thank you very much for the support!

i strive to make it correct and also digestible. it means a lot to hear this

Thank you so much for your support!

Thank you! I try very hard to be as clear as possible

Wow! Thank you for such a comprehensive overview on this subject. I take it you are in the industry yourself?

@@projectsinflight No, just have been using them for years, so learned why the Jfet and UJT have such a wide spread, while regular TTL is reasonably stable, and CMOS is pretty much immune in regards to threshold, really mostly being depending on gate oxide thickness for how it works.

While there are test patterns all across the wafer, it is infeasible to test anything electrically until the wafer is almost all the way through manufacturing. You need ohmic contacts from the semiconductor to the metal on top, which comes late in the process, because they have processing limitations and need to be at the very top of the wafer. So for example, the n++/p++ doping that is mentioned in the video could be used for a semiconductor-metal contact. But there is no way to check if the doping worked until after the wafer is completed. After ion implantation/for diffusion you need very high temperatures for a long time, then you need to apply the metal for the sc-metal contact and it's associated processing and only then could you test if the sc-metal contact works. And don't forget, each of the steps comes with photoresist application, patterning and cleaning!

CMOS functionally isn't too critical to doping, but not to say specs are unaffected: this is part of the reason why speed grades, input threshold, and output current, vary so much. The other part of the reason is they just DGAF and want wide guard bands on their process, lol. (You'll almost certainly never measure a 74HC or other gate in the wild that hits the datasheet worst-case limit say for propagation delay. Unfortunately, this is needlessly conservative when doing actual timing chains -- the one upside is, no one does that anymore, if you're doing logic, you're writing for an FPGA or ASIC... or building a hobby computer that you can test, select parts, and repair as needed.)

I appreciate it!!

i'm hoping i can also get the next video out soon

Actually I have been also working on DIY photoresist for months now. I'm not sure how close I am to a video, but I have at least one working formula. I'm also trying to recreate a novolak based resist from the 70s/80s. I will 100% make a video on all this when i am satisfied by my results!

@@projectsinflight Awesome, will be looking forward to that, as it would definitely be much easier to make the photoresist in the amounts that one actually needs, rather than buying a half a liter bottle and then 400ml of it goes bad two years later haha.

I'll be honest- the dopant I was able to make isn't perfect- it has to be stored at low temperature to have a decent shelf life. However, I think that it's the best i've seen so far. Maybe a real chemist can do better :)

thank you very much! i appreciate it

The next video will definitely contain lots of lab work!

while i love learning from youtube, i feel that education requires closing the loop on knowledge, meaning that you need a way to check that knowledge is actually being absorbed. I will say that i think there are better ways to close the loop than just taking tests though.

Yeah this could use more emphasis (if one want to get deeper into semiconductor physics). The remaining phosphor atom will be positively charged, which has huge effects - mostly creating pn-junctions, but also causing coulomb scattering.

that's wild! i have never witnessed that

Yup, silver and zinc as well. Hence the ~high school experiment plating a penny with zinc then heating it to brassify the surface -- you really are getting a brass layer, it's not just the thin layer of zinc oxidizing to an interference pattern or something. Silver (directly on copper) is semi-commonly used for PCB plating, but usually it disappears into the solder by the time diffusion would matter much. Along with tarnishing, it can be a concern in storage though. These also illustrate the Kirkendall effect, where the diffusion of atoms into each others' bulk reduces the density in each, creating voids. Well designed plating sequences (or diffusion schedules) avoid this in most cases, but there are industrially important applications where the voiding affects joint strength -- Sn-Cu diffusion (and intermetallic formation) in solder joints, for example.

thank for the support:)

Don't forget PN diodes as well!

For the longest time [growing up], I thought you could just take two blocks of the stuff, *now kiss .png* and boom, diode, transistor, whatever. (I tested, and wondered aloud, why wiring two diodes together doesn't make a transistor. Hey, I was... probably 9 at the time?) Sadly, all those early/easy/common references neglect to mention two things: 1. charges move by diffusion, and only travel so far before recombination dominates (in Si, a few µm, maybe 10 tops); 2. you need a metallurgical bond; a random misaligned crystal boundary with oxide inbetween, ain't gonna cut it. Astonishingly, reasonably-well-aligned crystals, and vacuum-clean and atomically-polished surfaces, can be created, and bonded metallurgically, so we can actually do this if need be(!), but it's a lot more work than rubbing two rocks together, hah. Equally surprising, it is still true that you can rub rocks together -- or more to the point, poke them with wires, "whiskers". Rough (read: non atomically polished) surface contact is dominated by asperities, where rough spots stick out enough to touch, give or take contact pressure and the elastic modulus of the material; this can happen with enough force to scrape off the oxides and get direct semiconductor contact (or a thin enough line of oxide, adsorbed air, and other surface contaminants, that charge carriers can tunnel through the barrier), and thus a diode is born. Transistors are harder: MOS is out of course, but BJTs can be made by microscopically-aligned point contacts. They also fairly early-on discovered you can fuse the whiskers into the chip, making a much more consistent, reliable, robust device; the 1N34 germanium diode is such an example, and the whisker (tungsten I think) is coated with a dopant so that when the device is assembled (glass sealed body) and activated (pulsed with near-fatal current), the point contact is melted and a PN junction formed. Or on a more mundane level, we can for example probe the carrier type of a semiconductor by simply poking probes at it, making sure they scratch the surface (ohmmeter reads ~finite), and then heating one or the other probe (or from that side of the wafer) to cause charges to move. Some 10s of mV are created in this way without too much trouble, so a pristine contact isn't required; the polarity vs. thermal gradient depends on the carrier type.

The very scary table you mean :P

thank you for watching :)

thank you for watching:)

I tried plain phosphoric acid but it really doesn't seem clean enough for my purposes. high concentrations leave a horrible residue (silicon phosphide maybe) on the surface that cannot be etched away

soooooooooooooon

To be fair one of them are percusors of chemical weapons.... No wonder it gets deadly real fast

that means a lot! thank you :)

I appreciate the compliment!

i've been contemplating a history-focused video

thanks! i appreciate the compliment

thank you- it means a lot :)

My impression was that it was only suitable for very shallow junctions and wasn't a good general purpose solution. To be completely honest though, I haven't really studied GILD all that much. It's possible I should consider it more closely

honestly i feel like my dream job would be to work for a company that manufactures or uses electron microscopes. i love SEMs

unfortunately, the diffusion step has to take place at ~1000C so the photo mask alone won't hold up at those temps. An oxide layer that has been patterned using a photomask will though

I hope to become even better in the future :)

I see what you did there :P

i'm very excited to show the progress:)

thank you very much too :)

I'm not 100% sure on this answer, but my understanding is that as long as the discontinuity between the metal and semiconductor is thin enough, the electrons tunnel through and basically ignore the junction.

I'm sure you can make it work somehow, but I definitely wouldn't trust my furnace to run at substantially higher than atmosphere. I'd hate for it to break again lol

hopefully some actual lab work ;)

i'm glad i was able to help :)

oh i can't wait to start talking about some of the chemistry i've learned about recently

have you heard of thermionic generators? Those might be similar to what you are looking for

i'm not sure i know what those are

@@projectsinflight Ah ok, then nevermind 😂 Pretty interesting though. When you wanna go into simulating a multi-electron system, you usually simplify a lot first. E.g. Hartree-Fock assumes that the electron in your material experiences not only the potential from the nucleus, but also a mean-field averaged over all electrons around it. But obviously, this is only an approximation and ignores direct electron-electron interactions (electronic correlations). These correlations occur a lot, e.g. for transition metal oxides

Oh i'm definitely continuing. Thnak you for your patience though- this last one took FOREVER to get right

to my knowledge i don't have a fiber laser but i appreciate the suggestion!

Oops my mistake! Good excuse for a new toy, though.

i'm going to cover that one on my next physics video

oh interesting- i've never seen it in person

I was considering adding it, mostly to come full circle on the "exposed core of the nuclear reactor" joke from earlier but it got cut because nuclear transmutation doping is really never used outside of a couple research experiments

i expect the next video will contain a very poor diode lol

absolutely! glad to make something interesting

thank you for watching :)

haha thank you very much

Me too! I hope to have it out in a few weeks

@@projectsinflight please man speedup 🥹❤️🤝💪💪💪

Do you know where to find good information on this topic? i've never heard of them

@@projectsinflight what do you think they are and how they work

thank you :) i do my best

glad you liked it :)

glad to be back:)

I see what you did there :P

if you're looking for information on energy storage in silicon you should read up on quantum wells and things like charge coupled devices. That being said, the energy density will likely never approach that of chemical batteries.

Yeah, I thought about including this to come full circle on the "exposed core of a nuclear reactor" joke but ended up cutting it because nuclear transmutation doping is basically not used outside of research labs

Thanks! I appreciate the support

i see what you did there

I appreciate it :)

You can! In fact, back in the day they'd actually make electrical contacts by applying a tiny bead of aluminum and putting it into a furnace. Some aluminum would diffuse into the chip, making a doped region and then the rest would stay behind on the surface as an electrical contact.

doesnt that also give you a shottky connection?

@@suncrafterspielt9479 It depends on the doping amount. If you have a metal-semiconductor junction where the silicon is lightly doped you get a diode, if it's heavily doped you get an ohmic contact In the process i described in the first comment it results in a heavily doped region, so it should make a non-rectifying junction

Thanks a lot for all the explanation :D

👍

thanks!

probably one more week

​@@projectsinflight **1.06 weeks later** Sup, you got da spinny go go dope? I'm on a quest to make sand think and need the good stuff 😂

@@cgarzs I'm taking applications for unpaid video editors if you'd like to apply

@@projectsinflight I wouldn't pay me for my editing skills either lol. I'd just upload the raw footage with a single incoherent comment for context and call it a day 😆

Thank you for your patience- the next video took quite a while to figure out

Thank you!

i was thinking about it but it needs to be kept very cold or else it goes bad in a week. haven't figured a way around that one yet

you know i've always wondered how it would work trying to use diamond as a semiconductor...

I see someone is a fan of neutron activation doping ;)

provide me with links to relevant information and i can include a correction on the video

It's a good question! For starters, I'm not in the industry, though I did a bit of related lab work back in college. In industry the demands on their processes are huge- they are desperately trying to make chips with billions of individual transistors, all while trying to maintain the highest possible yield (number of good chips vs defective chips off the assembly line). These goals mean that if you have the choice between a safe chemical that is effective, and a toxic, carcinogenic, explosive, and otherwise extremely scary chemical that is *slightly* more effective, you'll choose the more dangerous, but more effective route every time.

thanks :)

electron mobility is pretty high so i'd say... probably

haha ;)

fortunately, the elements relevant to the video have been known about for at least a couple hundred years

Some people prefer more explanation than others, and I think it’s nice that this video takes the time to teach.

I was going to make it all one video, but i decided that if i went straight to the lab work then to keep the video length less than 30m id have to gloss over a lot of info that i felt was necessary to lay out in detail first. i am curious about the rice hull and glycol thing, if you have a source you'd like to share