Homemade aluminium semiconductors
If you find any of this text dry then just skip down to the pictures
Previously I've covered making copper diodes at home. Unfortunately this type of diode is almost as unwieldy and fiddly as using a cat-and-whisker detector. It's a far cry from something that would allow me to easily make diode matrices on a PCB.
Traditionally we see semiconductors as being very expensive and difficult to make. Who could blame us? They all come from expensive cleanroom fabs that use multi-million dollar equipment. Heating a pizza in one of those chip ovens is industrial sabotage. How could you ever hope to get close to that level of control and precision at home?
A special thankyou to everyone who commented on the Hackaday article covering my last post about this. A lot of you suggested great ideas to investigate and I have been going down a few of those paths because of you.
Copper diodes look like a dead end
Everything I've read so far suggest that copper diodes were commercially made by heating copper up to around 800 to 1000 celcius. This would build up a couple of oxide layers and you would chemically strip the undesired outer layer.
Although I can easily get copper to that temperature, getting it to that stage whilst it's on some sort of board or PCB is very difficult. I don't have many materials at hand that can survive 1000 degrees, let alone ones that would be suitable as a base material and also easy for other people to get their hands on.
Whilst researching copper diodes I can across this interesting document:
THE COPPER OXIDE RECTIFIER
By Thomas Mark Cuff, Engineering Department of Temple University
"Luck sometimes visits a fool, but never sits down with him."
That quote bodes ominously.
Anyway, this doc covers some history of the copper diode, including some discussion about how copper sourced from Chile seems to work better. But in its historical overview it mentions 'wet type rectifiers':
At its simplest, a wet type rectifier consisted of, for example an electrolytic cell composed of an aluminum plate and a lead plate dipped into a solution of water and borax. When the aluminum electrode was positive (anode), oxygen was evolved at its surface producing a thickening of the alumina (Al2O3) passivation layer, which due to its insulating nature would prevent any current from flowing; the passivation layer rapidly dissolved when the aluminum electrode was negative (cathode), thus allowing the current to pass unimpeded through the cell. It was this growth and dissolution of the insulating alumina layer which gave rise to rectification in the wet type rectifiers.
And so the diversion begins.
The paragraph continues:
Though wet type rectifiers are long gone, their direct descendants are still with us: aluminum electrolytic capacitors, which depend for their proper functioning on the electrolytically 'formed' insulating layer of alumina on their plates. These capacitors are, of course, polarity sensitive, if put in the circuit backwards the alumina film will dissolve resulting in the shorting out and possible destruction of the capacitor, due to overheating as a direct result of increased current flow.
Indeed, this makes sense.
If you use an electrolytic capacitor forwards: it slowly starts to block your current flow as it charges up. A tiny electrolytic capacitor blocks it almost right away.
If you use an electrolytic capacitor backwards: it looks like a short.
If you want to use one both ways: you wire two in series, facing opposite directions. 'Non polarised electrolytic capacitors' do this.
Baking powder versus Baking soda
Some sources suggest that you can use baking soda or baking powder instead of borax as the electrolyte.
I have a hunch that 'baking soda' and 'baking powder' are sometimes used interchangeably and sometimes used to mean different things. Later I'll show some investigations into using both of them.
When I use the terms on this page:
- baking soda = straight sodium bicarbonate (basic)
- baking powder = a mixture of soda bicarb and an acid powder (and possibly other things)
Making and testing
To make an electrolytic capacitor (read: rectifier) you need:
- an electrolyte
- another conductor on the other side
Ideally you don't want to use aluminium on the other side as then your device will develop diode characteristic in both directions. From what I've read you can use most metals. I suspect that actual aluminium capacitor manufacturers use different metals or at least different aluminium alloys on each side, but I may be incorrect.
In my case I used:
Aluminium: A tiny strip of aluminium foil cut with scissors.
Electrolyte: Water and baking powder mixed together. I soaked a small piece of paper towel in this solution and use it to both hold the electrolyte and to seperate the two metals.
Other conductor: copper pads on a copper PCB.
I also mixed up some conductive glue using powdered graphite and some clay. The clay was plain white stoneware, but anything should be fine. The graphite was from a bottle (you can get this for lubricating locks), but you may have success rubbing a pencil on a file instead.
Use as much graphite as possible and as little clay as possible. If you don't have enough graphite in your mix then your joints will measure hundreds of kilo-ohms. I was able to get this down to the few Kohm region. My clay, even when wet, had very poor conductivity on its own (which was useful for the paste diodes I show later).
First up I prepared pads on a bit of copper-clad board to make the diodes on. They were seperated by cutting lines with a hacksaw.
I then starting laying the components onto this PCB:
Here you can see the little pieces of soaked paper towel and the aluminium strips. I found that sitting the aluminium strips onto the towel led to poor contact/adhesion. If you fold the towel over the aluminium then it holds together much better.
On the other side of the diode I simply needed to connect the aluminium foil to the pad. This is where the conductive glue came in:
The diode will work like this, but for it to last more than a short time you need to stop the electrolyte drying out. I originally tried covering everything in hot-glue:
Unfortunately this didn't work well:
- Sometimes I would 'drag' parts of the diode with the glue, destroying them
- I didn't do a good job sealing things, so the diodes dried out anyway.
I need to find a better way of doing this.
The next step is to 'treat' the diode. By passing DC through the diode in the direction that it should be blocked you build up an aluminium oxide layer. Effectively you pre-charge the capacitor. This is useful for testing if the diode works -- ideally the diode pre-treats itself once in-circuit.
I used 36V and a 1K resistor to do this step. You put the positive terminal of your power supply onto the aluminium side of your diode.
Finally I hooked the diodes up to a curve tracer. I used the circuit from my previous post. If you don't understand these next diagrams then that page will help explain things.
Lo and behold:
5V per horizontal division, 5mA per vertical division. As you can see the turn on voltage was about 5V, which is much higher than the usual silicon 0.7-ish volts.
Unfortunately this behaviour rapidly degraded. Over time the reverse breakdown voltage (seen as a little turn on the right end of the graph above) slowly moved inwards. After a few minutes you would start seeing this:
The diode is still a tiny bit better forwards than backwards, but it's by less than a factor of two, which is not good.
When the diodes were useless a day later (from drying out) I pulled the hot glue off. This picture reminds me of when people tear chips apart when removing heatsinks:
I also didn't have any success using 'baking soda' instead of 'baking powder'. The diode would not block current and the copper would leech into the electrolyte:
What is this baking powder anyway?
In Australia the ingredients of all food products are listed on their packaging (with some exceptions, such as single-ingredient products). Importantly they are ordered depending on how much is used in the product.
In other words: this baking powder is mostly made out of rice flour. Time for some investigation!
Testing a variety of electrolytes
Process: try different metal combinations and current directions, see which electrolytes block current (ie act like a capacitor/diode). Again I was using 36V through a 1K resistor.
I didn't have luck with a curve tracer in these tests: the capacitance was too large, giving me curves like this that you had to mentally average to get rid of the loops:
As such it appears these diodes are too slow to be greatly useful at 50hz operation. YMMV, especially with different diode sizes and loads.
I tested all of the chemicals shown in the photo above:
- nothing but water
- table salt
- rice flour
- baking powder
- baking soda
The results showed that baking powder, soda and borax all worked. Salt acted like an almost dead short (compared to the 1K resistor) and everything else was unexciting.
I tried using aluminium, copper and graphite as the second terminal across all of these electrolytes. Using graphite or copper as the second electrode both had similiar results. Bubbles always seemed to emanate from the negative electrode, so the electrolyte (and possible electrode) is being consumed during use, which may cause issues later.
But why was baking soda all of sudden working in this test when earlier it did not?
I tried to repeat this test by mixing a new batch of water and bicarb soda.
I tried again with 'de-ionised' water instead of tap water. Same result.
Current would start to drop as if the setup were acting as a diode/capacitor, but it would give up after a short while and start rising again. If I left the setup unconnected for a few minutes then it would work again for a short while before failing.
It turns out that I have to mix soda bicarb and water, put the aluminium in and then leave it sitting for a day. After this it seems to work well. I did this accidentally in earlier tests where it worked, and I've confirmed it again today. I'm not sure what's happening here: perhaps other interfering materials in the water are slowly precipitating out or otherwise escaping/changing. Perhaps the aluminium is dissolving into the electrolyte and saturating it.
Anyway, it's something to try if you have similar issues.
Manufacturing the paper-seperated diodes was fiddly and error prone. It required tweezers and the tiniest shake of the hands would ruin the diode.
Ideally I want something that can be 'dabbed' or 'printed' on to make diodes. Ie not require fiddly assembly.
Something like this:
The white paste is a mixture of clay and baking powder. The grey is a mixture of clay and aluminium filings. I rubbed a piece of aluminium against a file to get these. The red is just electrical tape to make the gap wider (the copper pads are seperated beneath it, as per earlier).
The performance curves were similar to earlier and did not seem to degrade as easily, but the overall resistance was much higher. I'm going to investigate further with these to see if I can get them to work better.
Also: these diodes still need to stay wet to work. Putting a drop of water back on a dry one does seem to work, but you have to be careful not to put too much (and bridge the copper pads).
I'm not quite sure what to make of things. I've gone much further than I expected. Although I've not exactly discovered anything new, I've had some success with making up my own manufacture methods. Soon I hope to make a diode matrix for 7-segment display driving. That would be fun to achieve.
Obligatory "why don't you just buy diodes, they're so cheap?". Because that's not fun. And you don't learn anything new or discover anything new.
Have some fun, and post if you have some thoughts, ideas or if you give things a go. Unlike last time: no login is required to post.