Making a cloud chamber to see radioactivity


Several years ago I was intrigued by a strange table at the Palais de la découverte in Paris, which was showing randomly moving lines supposed t be traces of particles falling on it. By then I found no explanation about it.

Later I learned that this is called a cloud chamber, which is quite well explained here.  These devices are not very hard to build and I happened to have all the required parts, so I felt urged to build mine.


The principle consists in evaporating alcohol in a closed but transparent jar, above a very cold metal plate. This plate being between -20C and -40C, it condensates a small part of the alcohol vapor showing a thin layer of tiny droplets which move quite fast. By lighting them horizontally it's possible to see them moving. When a charged particle travels through the cloud, it ionizes some of the droplets, increasing their density and showing a brighter track for a second or two.

First attempt

I first tried to stack several Peltier elements in order to increase the temperature gradient between the hot and cold side. This is the most difficult thing in fact. Indeed a typical Peltier module will consume around twice the heat it extracts, thus it will emit 3 times the power it extracts from the cold side. If you stack modules with less than 3 times the Qc rating (extracted power), you will not manage to cool the cold side enough. At 3 times, the colder one is almost useless as the hot one will suck just the extracted power. I found that multiplying the power by roughly 4 is what allows to provide the best gradients but it is hard to achieve.

The second problem is that with two layers you can roughly lower the temperature 40C below ambiant temperature. At 20C ambient, it just gives -20C. So I really wanted to stack 3 layers, but this means producing 9 to 16 watts for 1 watt extacted. However on the cold side we don't produce heat in the cloud chamber design. So we can afford to have little extaction power on the cold plate if the insulation is good enough.

I first tried by stacking 3 cheap Peltier modules (12706, 6amps under 12V) and adjusting their voltages. I installed a small anodized aluminum plate on top of last one and an old server fan on the hot side. I managed to reach -37C, to see a small cloud but couldn't notice anything in it.

Looking back at various deisgn notes, I found that a high voltage gradient is usually needed (positive at the top, negative at the bottom).  Some report that it also works (though less efficiently) without the high voltage source. Others indicate that the alcohol needs to be heated and that the top of the chamber must be sealed. And I didn't have any radioactive source (at least nothing I was certain was radioactive).

Thus I decided to order new parts.

Second attempt

For the second attempt I decided to order higher power Peltier modules. I got a 12715 (15 amps) and a 12710 (10 amps) to experiment a little bit. I also ordered a small set of tiny uranium-doped pearls and a fire alarm module using an americium 241 source, to have something to test my device.

I decided to use a larger cold plate. I found a transparent box which had been used to present a toy car, which had the good taste of being black at the bottom, slightly longter than two Peltier modules and larger than one, and with very clear sides. I installed a metal grid at the top, and used thick wires as supports, passing through the plastic base to connect to the high voltage source. The bottom plate is now about 8cm x 4cm which covers two Peltier.

I've cut thick copper plates (2mm) to arrange two modules per layer. It turns out that copper is too soft and very hard to keep flat once cut.  And pressing on the sides is often enough to bend it, so I ended with a few bent copper plates which were not making a good contact. Then I figured that my fan didn't have a good contact over 8cm. I spent hours filing it to try to make it flatter, and eventually it worked. Thermal paste is also a problem. The silvery ones are not sticky at all and can let the plates go off very easily. The white compound made for transistors is way better but doesn't conduct as much. In the end I found that mixing the two gave a very good result, it becomes a gray sticky paste with high thermal conductivity and enough viscosity to keep the modules in place.

I don't count how many times I had to disassemble and reassemble everything, using large setups is a huge pain as you cannot even afford 0.1mm difference between two extremities.

Then I found that it was not possible to go below -13C with this setup, whatever voltages I used on the various layers. In fact the 12715+12710 modules at the bottom (hot side) were producing too much heat for the server fan to cool down and I noticed that the top of the fan with burning hot while the air out of it was just warm.

So I decided to replace the heatsink + fan assembly for an old 1U copper heatsink made for a power-hungry Pentium4, and started to stack as many fans as needed to blow into it strongly enough. In order to keep the top in place, I'll use thick steel plates on the sides.

These ones were assembled using threaded rods which also serve as feet to keep the assembly away from the table :

So I tried 1, then 2, then 3 fans. They are 12V/0.25A each (3W), but I power them under 19V (~7.5W). Normally fans do not blow fast when you stack them like this, but here it's different, the heatsink's fins are extremely thin and the air pressure between the fan and the heatsink is high, so a lot of power is needed to further increase this pressure. With a single fan top of the heatsink was at around 44C. With two it was at 40 and with three it's around 37. So that's 7 degrees won :

For the Peltier modules, after lors of experimentation, I ended up with this design :
  • at the top (cold plate) : a single 12706 module, powered under 4V through a DC-DC converter.
  • just below it, two 12706 modules in series, powered under 12V. This results in roughly 6V/3A per module, so this layer is about 3 times stronger than the first one.
  • and below, the third layer is made of the 12715 and the 12710 modules in parallel, totaling 25A under 12V, to suck the power produced by upper layers. Note, for a better design I should have opted for two 12715.

The plastic box is then installed on top of this. There needs to have enough clearance below it so that it can press the plate to keep the modules in place. It is important not to screw it too strongly so that it doesn't bend the upper plate :

Note that at this point there is no lighting installed in the box. I'll simply use a handheld torch for now. For testing the temperature initially didn't want to go below -13C again. I figured that some of the hot air from the fan managed to go below the cold plate and to heat it. I then installed some polystyrene foam between the bottom layer and the top layer and it went down to -32C this time. Due to the heavy copper plates it takes about 5 minutes to reach the lowest temperature. I photographed it inside under low light conditions to see better (no alcohol yet, only the moist in air freezes) :

In order to start to see the cloud, I let it cool down, then heated some alcohol in the microwave oven. It's easy and allows it to quickly evaporate, you just have to watch and stop heating when it boils. I'm using 95% ethanol, which works pretty well. Others mention the need for isopropyl alcohol but it is much harder to find and not necessary. I've placed some absorbing paper on the top grid and just poured some hot alcohol on it. I quickly closed before it evaporated and connected the high voltage supply.  First some alcohol started to condensate at the bottom and to dilute with the water ice which had already formed there, and at some point while lighting with a torch, droplets were clearly visible at the bottom, indicating that the cloud had formed. I didn't notice the effect of the high voltage and am stil having doubts about its effectiveness; the cloud has the exact same aspect with and without.

Very quickly I started to see thick tracks spontaneously appear in it! It worked! The straight ones are made of Alpha radiation (42He atoms ejected as the result of the transmutation of Americium 241 into Neptunium 237). The broken lines are made of electrons or positrons (Beta radiation). On the images below, the left one shows the chamber with no visible track and the americium module installed on the right. The middle one shows a track in the middle coming from the americium module. The right photo shows another track at the top (less visible).

I uploaded a video here.

Lessons learned

Peltier modules advertise currents that don't match reality. The cheapest 6A ones (12706) really draw only 2.5A approximately. In fact it depends a lot on the temperature gradient between the two sides. Some of them are unmatched so it's not easy to connect them in series as they will not react similarly. Also, connecting them in series may require to increase their voltage if they are not strong enough. I think instead I'll buy some high-power DC-DC buck regulators and install each stage on a separate voltage so that I can control them individually. I noticed that it's possible to significantly reduce the amount of power once the plate is cold. Typically instead of 12V, the setup continues to work at 8, which is about 2.25 times less power, and makes the fans more efficient. In fact, what matters is to maintain the most optimal power levels at each stage. One must start from the coldest side, see how far it's possible to go without heating the next side too much, then start the second side, then the third one.

I miserably failed at using two modules for the coldest side. I think that the difficulty is that some of them are not as good as others, and that since it is very difficult to make them work at very low temperature, the gradient remains the one of the poorest of the two. If one module requires 2 watts to reach a low temperature and the other one requires 3 watts, with two modules in parallel I'll emit 5 watts which make it way harder to cool down by the next level, while 4 would have been sufficient with a set of identical modules. I don't know if larger modules exist. These ones are 4cm wide, it would be nice to have larger ones, it would simplify the setup. The 15A ones (12715) are theorically more efficient than the 6A ones as they correspond to 2.5 of them placed in parallel, and have a lower resistance. Thus in theory by running them at a lower voltage and under the same current, they should produce less heat (less Joule effect). So I think it's better to use higher power modules and run them on a regulator than using several low-power ones in parallel.

Since not much heat has to be extracted, a single module is sufficient. It is important however to put some polystyrene foam between the plates, especially the top one, to prevent radiated power from the lowest ones to heat it up. I intuitively think that this layer has to be thin enough not to touch the two sides in parallel, thus it becomes a polystyrene+air insulator.

Fingerprints (especially those with thermal grease) tend to mark the anodised aluminum plate a lot. I'll probably have to clean it up several times again as it's harder for the cloud to form on top of these traces. Running the device "dry" (no alcohol) lets air moist freeze on the plate and reveals the traces, so it's easier to spot them.

Future improvements

I need to install a led strip around it to avoid having to hold a torch. I'd also like to experiment a bit with higher voltages to see if they really produce something. I've read some reports that high enough voltages increase the probability to see tracks, though it wasn't obvious here. Anyway it was an awesome experience which was worth all the difficulties I went through while building this device! I'll probably have to choose different Peltier elements.


  1. Great work and very useful information! It is fascinating to see how people go different routes when building the same device. My son and I built a Cloud Chamber but we opted for liquid cooling of the peltier elements and using lower power elements for the 3 stage (12709). I was considering going to 12710 or 12715 but seeing your work I dont think it will make a huge difference in my setup. My build is at blog.kotarak.net - check it out!

    1. Oh excellent, thanks for the link! I'm seeing that we basically use the same lighting method though :-)

      I bought dual-layered peltier elements but have not tested them yet. They should make the whole thing colder.

      Some people also simply use dry ice at -70°C and it seems to give them excellent results, plus it easily allows to have a large area. But I don't have dry ice and I prefer something that can be instantly powered on without depending on stuff I'd have to buy from a remote shop.