# Can Peltier devices be cascaded?

Can Peltier devices be cascaded to create a bigger temperature difference? Like mounting one on top of anther one to increase the maximum difference from 60 degrees C to 120 degrees.

Yes, and this is regularly done. However, there are limits to what you can achieve, based both on the limits of the individual devices (minimum and maximum temperature) and effects such as the total thermal resistance through the stack. Eventually you get to the point at which the "reverse leakage" of heat through the stack (which rises with the end-to-end temperature difference) equals the stack's ability to remove heat.

Another problem is the relative inefficiency of Peltier devices. Typically the heat flux coming out of the hot side of each device is on the order of 3 to 5 times the heat going into the cold side. As you stack devices up, each one needs to be that much larger than the previous one, leading to problems with sheer size (which also gets back to the heat leakage problem).

• Would this issue still be present when effectively using Peltiers to heat up and not cool down? – joaocandre Apr 18 '14 at 16:36
• I guess it's goes both ways. It's just related to the module's heat conductivity... – Tomáš Zato - Reinstate Monica Jul 21 '14 at 0:06
• For pumping heat the reverse leakage will not matter as much but you will reach the junction temperature fairly soon with multiple stages. It seem uneconomic to use stacked Peltier elements for heating (a resistor is much simpler) unless you have some very critical energy savings to achieve. You can save perhaps 20% of the energy by scavenging from the environment. – KalleMP Nov 25 '16 at 7:00

Sure, however due to the miserable efficiency it's typical to stack the in increasing size, sort of like rocket stages, so the fattest is dealing with the heat flow from all the others.

Photo from here.

They definitely can be cascaded, but the problem is that warm stage could have much more heat transfer capacity than cold.

AFAIK most effective thermoelectrics have transfer factor of ~100%, mean, they consume energy and produce heat of 1 W per 1 W transferred from cold side (compressor based fridges have about 300%, they transfer 3 W heat per 1 W power).

Say you need to transfer about 1 W of heat from you device. Then coldest stage could produce 2 W of heat at its hot end, and all its heat should be transferred by next stage. Next stage will produce 4 W of heat. Then 8 W and so on.

Cascaded peltiers should look like this:

• I wonder what is the device pictured used for - do you have any more data on it? It looks rather pretty. – Richard the Spacecat Jul 27 '18 at 15:02
• Low noise photodiodes for APD single-photon counters. Also thermal IR sensors/imagers. I found a list: tec-microsystems.com/about-us.html – wbeaty Jul 27 '18 at 18:38

Yes. You can stage multiple single stage Peltiers, if due regard is given to the electrical and thermal flows. You will see that multistage devices usually have decreasing physical areas for the colder stages. This is because you have a decreasing amount of "coolth" available at each successive stage as the hotter stages before them have to pump both the thermal energy from the colder stages plus the electrical resistive losses from the colder stages.

Due to the low efficiency of Peltier coolers wrt electrical input a cold stage must be operated at substantially lower electrical input that the hotter stage that is cooling it. It is easy to swamp the hotter stage with thermal energy from the DC input from the colder stage and get no net cooling at all.

• I am mostly interest in the warmth generated, so if I stacked two cheap 60W peltier devices such as [these][1], I would not be able to achieve twice the dT, but I would still reach around +100C right? [1]: ebay.com/itm/… – joaocandre Apr 15 '14 at 13:02
• @joaocandre - About now you need to tell us what you REALLY want to achieve and try to avoid the death by 1000 increments when the actual requirement and what is stated do not match well. What are you trying to achieve? Are you trying to make a heat pump, a cooler, a ... ? What sort of desired dT. What ambient and output target temperatures? Why? – Russell McMahon Apr 16 '14 at 12:43

Directly stacking Peltier modules is problematic in practice. The heat-sinking required is substantial. You can think of a Peltier serial array (stacked) within a system as a machine that must must be 'started'.

If the heat sinking is too substantial, it takes forever to start the heating/cooling. This is easily compensated by using a fan with the heat sink, and then throttling the fan low on startup.

Though I fail to understand the advantage of Peltier based heating, other than in a system that switches between heating and cooling for the same task.

Resistive elements are more durable and easily controlled than Peltiers for heating, because they can be hard-cycled many times.

The design I used for multiple Peltier modules stacked was one 12706 between a heat sink/fan on the output side and a finished copper bar twice the width of the 12706, on the discharge.

On the other side of the copper bar were (2) 12706s in parallel, mechanically, and a heavy, aluminum heat sink/fan on the final discharge side.

The individual Peltier (TEC) elements were wired in parallel. I drove the parallel array of 12706s with a maximum 15ADC, 12VDC, RTD-disciplined, linear PSU, constant voltage.

Linear PSUs are inefficient in and of themselves. So, RTD-disciplined SMPS (> 90% efficiency) is a more efficient option.

That system was for cooling (attained -12C at room-temperature ambient), but if you reverse it, it will work for heating. Peltier elements must not be heated above the temperature of the solder used to make them. Careless or inexperienced experimentation can easily bring this about.

You just want to ensure (2) things: that you don't sink too much heat off the hot side, because the heat transfer depends on the temperature difference of the two sides. That property of TEC modules has idiosyncratic limitations.

If the hot side isn't hot enough, the system won't transfer heat, and the power draw will be low. And also that the heat transfer doesn't turn parasitic and run out the cold side, so the entire array is just a heater. That can melt the solder in the TEC (Peltier) module.

I found the most useful spec on a TEC module is the optimal rated ranges of temperature on the hot and cold sides. Everything else but the electrical input can be derived by experiment. But if you try to obtain the specified deltaT using incorrect high and low temperature, you might not obtain the full heat-transfer capacity of the module.

Much of the advantage obtained with quality TEC modules is that they operate with the rated temperature differential shifted lower. 66C delta can be 44C-100C, or 0C-66C.

Not all deltaT >= 66C rated TEC modules will operate well at delta 0C-66C, or lower. They may give the greatest heat transfer at delta 44C-100C. The cooler the cold side gets the more desirable the system, usually.

It is also required that thermal-transfer interface compound be applied between the TEC modules and what they interface to. No TEC module interfaces directly to the atmosphere. There is always something on either side of the Peltier modules.

I was 'unable' to obtain satisfactory results directly stacking a 12712 onto the hot side of a 12706.

Around 2009 I prepared one such cryocooler, using a 3-stage Peltier stack and largest sold air-cooled PC videocard cooler with 7 heat-pipes (You may search publication for ROFLEX iodine instrument).

Later the client wished to construct more copies going around me and asked to factory of Peltier plates what air cooler he ought to use. Answer I was proud to hear was - with our platelets is is completely impossible to cool down anything, using any type of the air-cooler. Just the problem at Peltier is not a heat flow as such, but a heat flow intensity over radiator contact-place square centimeter. Platelets have a rather small size, about 4x4 or 2x2 cm, thus the 100W heatflow is there more than much.

Actually, 3-cascade platelets in my case gave a 116 C difference betw endplates, what is near to the theoretical boarder, thus I got to produce stable minus 45 C at tropical climate.

This year I need to get even more, the -100C for 1 cm3 by non-water cooled radiator when +50C air would be goal on it. For a while I am not sure is it possible at all.

Im writing this to assure that -45 C is really possible, but not much deeper. Theory says that 4-rth platelet over three will damage the process instead of boost it.

Yes, tiny devices are actually sold for this purpose. You can even buy ready made stacks with the appropriate materials on each stage so that the heat saturation problem is mitigated (if I recall they use a higher efficiency process at the top by substituting Sb into the BiTe) I have plans to try and cool a sample of SH21Pd97? under pressure down to near cryogenic temperatures and see if the resistance drops suddenly, and duplicate the same experiment with Bi-2223 as a control with a few tweaks later that might boost Tc by up to 20%. Possibly even laser enhancement via tuned IR laser.