Thermoelectric Cooling | Hacker News Reader

Thermoelectric Cooling

thermoelectricsolutions.com

I recently discovered that there are theorized devices, called phonovoltaic cells, that also convert heat energy to electric power [1, 2].

In a PHOTOvoltaic cell, incoming photons give electrons the energy that eventually turns into a current. Like the photon, a phonon is also a packet of energy, but in the form of a moving vibration inside a solid. In a PHONOvoltaic cell a phonon gives its energy to electrons to create a current.

Unfortunately, we have been unable to find materials with the right thermodynamic properties to create these devices. But there is progress on showing that graphene could be it.

I guess that a phonovoltaic cell, like a photovoltaic cell, would have efficiencies much higher than TECs (5-8%), and so would make a lot more economic sense.

[1] https://en.wikipedia.org/wiki/Phonovoltaic

[2] https://pubs.aip.org/aip/apr/article/6/2/021305/570297

Would this require a temperature gradient to function? My understanding is that a device that converts heat into power would violate thermodynamics.

A photovoltaic cell also requires a temperature gradient. In the case of a solar panel, it's the high temperature of the sun that shifts its blackbody spectrum into a range where the cell can generate electricity. But if you were to heat the PV cell to that same temperature as the surface of the sun (somehow without melting it), it would glow and radiate away just as much light as it absorbs from the sun, rather than converting any sunlight into usable electricity.

Likewise the phonovoltaic would need to operate on a phonon spectrum that is shifted away from that of the material's own temperature, in order to not violate the laws of thermodynamics.

> But if you were to heat the PV cell to that same temperature as the surface of the sun [...]

Much easier to replace the sun by some other source of EM-waves of suitable wave length and put it into a freezer, below the temperature of the cell.

How does the wave poking some electron to a higher state remember the temperature of its source?

Good question. Temperature is a funny thing. When a system is far from equilibrium, the notion of temperature becomes a little unfamiliar.

The thermodynamic temperature of the radiation is connected to the entropy of its power spectrum. LEDs and especially lasers emit very low-entropy light, which can be focused and heat a surface up to very high temperatures. (Sunlight focused onto a surface cannot heat that surface above the temperature of the sun). Thermodynamically speaking, a low-entropy power source like a laser -- and whatever is driving it -- must have a very high exergy, which is equivalent to behaving like a high temperature heat source, even though it might feel cold to the touch.

Some more details here: https://en.wikipedia.org/wiki/Exergy#Quality_of_energy_types

Stored electricity's equivalence to a high-temperature heat source is one of the things that makes it so useful. It's intrinsically connected to why it takes a lot of low-grade heat to produce a small amount of electricity in the first place, and also why electric furnaces can produce such high temperatures. So while a battery can drive an LED that shines on a PV panel that generates power with everything feeling equally warm to the touch, a temperature gradient is still necessary; it's just been moved outside the boundaries of the system, to the process that distilled the entropy out of the energy that became the battery's stored charge. We can reversibly recreate this temperature gradient by driving a Carnot engine with that battery, instead of a laser.

> How does the wave poking some electron to a higher state remember the temperature of its source?

The black-body spectrum of the source determines the amount of waves it outputs at each frequency, and in turn is determined by the temperature.

I think semiconductors (such as most solar cells) do not act as ideal black bodies. Thermal excitation (what temperature required?) can move electrons into the conduction band and when the electron relaxes it emits a photon corresponding to the band gap.

I looked for a good article, but didn't find anything. There was one relevant comment in the first article I skimmed that said "semiconductors do not behave remotely close to a blackbody": sounds authoritative but the comment was on stack-overflow so hard to judge its correctness!

Oh yeah, I was talking in the context of the source being the sun. Of course other sources vary.

Right. The context you were answering is jbay's comment that "if you were to heat the PV cell to that same temperature as the surface of the sun".

Sorry.

That is fascinating, thank you!

Yes indeed. There is quite a lot of discussion on the wiki page about what constraints that laws of thermodynamics place on such a device.

> My understanding is that a device that converts heat into power would violate thermodynamics.

isn't a coal burning steam engine converting heat into power? perhaps the violation applies if requiring efficiency in the conversion?

Let's be precise. When we say "convert heat to power", what we mean is actually "if we have some energy flowing from a heat source to a heat sink, we can extract a percentage of that energy as useful work, like turning a steam turbine to produce electricity".

Carnot proved in 1824 that the maximum theoretical efficiency is purely a function of the hot and cold temperature (in Kelvin):

Eff = (T_hot - T_cold)/T_hot

So for instance a thermoelectric generator operating between 25 °C and 5 °C has a maximum efficiency of

Eff = (298 - 278) / 298 = 6.7%

> So for instance a thermoelectric generator operating between 25 °C and 5 °C has a maximum efficiency of

> Eff = (298 - 278) / 298 = 6.7%

Does that mean a thermoelectric generator operating at a minimum of 0K has a maximum efficiency of

Eff = (N - 0) / N = 100%?

Yes, although in practice you will have issues sinking heat at 0K without using additional energy to do so, since that's below the radiation sink temperature of space (~2.73K).

Or just (2*N-N)/N.

According to the formula you'd have to divide by 2*N.

my mistake!

Steam engines operate off the temperature gradient - the water (cool stuff) is turned into steam (hot stuff), which expands 16x and creates a huge pressure spike that pushes the piston/turbine.

> steam (hot stuff), which expands 16x

1600 times. Gas at room temp occupies 24 dm3/mole.

> My understanding is that a device that converts heat into power would violate thermodynamics.

I'm probably missing the point entirely, but aren't we doing this already with Thermocouples as used in RTGs?

There are a number of different effects that turn heat into electric power. The review article I linked in the opening comment discusses all of them.

The website appears down. I think I found it on the Wayback Machine: https://web.archive.org/web/20240106121752/https://thermoele...

These things are very exciting. Flat piece of ceramic. Put a voltage across it and you get a temperature gradient, put a temperature gradient across it and you get a voltage.

I think there's a niche application to computer cooling in making watercooled machines smaller. The reasoning goes something like radiators work really well at high temperatures - say 75C to keep the pumps from dying - and CPUs work better when they're colder.

Watercooling has a surface area problem - the water/temperature delta is a linear function of radiator surface area. Going from ten degrees to five involves lots more heat exchanger or very loud fans. Tricky if you're carrying the box.

I think there is a point in the design surface for burning lots of electricity to hold the CPU somewhere around (maybe slightly below) room temperature while moving that heat and the extra power from the peltiers into a water loop at well over room temperature. Maybe a 40C delta between water and CPU, which probably involves stacking peltiers for even lower electrical efficiency.

The point would be a smallish (like LAN party size) self contained watercooled box with terrible electrical efficiency and excellent thermal density. Unfortunately I don't actually have any sort of use for that so never checked the curve intersection holds up in reality.

I have used these for several projects. If you need a few W of cooling power to drop some material below ambient temperature, TECs can be a great choice. However, you need to now dump about 5-10X the heat somewhere else.

If you need to just cool (not below ambient) or have a lot of heat to move, this is not the right device.

For large stuff, a heat pump or typical compressor based AC unit are much more suitable, and you can move about 1500W for around $150 unit cost.

TEC is an absolutely awful choice most of the time. It's highly inefficent, requiring huge amounts of energy for a modest change in temperature.

That said, they are an absolute god send for some situations. They can be made way smaller than a compressor, so if size is your primarily concern, you can make tiny refrigerators that chew through power, but can keep a soda or medical vial cold.

They're also absolutely essential when vibrations MUST be kept to a minimum. Almost all astrophotography is done with cooled cameras which must not vibrate. My astrophotography camera has a pitiful micro 4/3s sensor it can cool 35C below ambient for the low low cost of 36 watts. 36 watts to cool 1/3 of a square inch. But it doesn't vibrate!

Tried to use one of these in a space application once: cold side on the sensor, hot side attached directly to a radiator. Unfortunately the thermal conductance [W/°C] from the hot side back to the cold side was high enough (relative to my radiator's thermal conductance to space) that I needed to pump not only the sensor's heat load, but a small fraction of the Peltier's heat as well, which ballooned the electrical power required. Not worth spending that much power to buy an increase in SNR from a cooler sensor.

Many CCD cameras used for astronomy use a TEC stack to cool the sensor. Separating the cold and the hot side is paramount. Haven't used it in space, but I'd probably use a vapor phase heat pipe to move the heat as far away as possible before radiating the rest of the heat away.

That would have helped a lot! Using a high-conductance device to transport heat away from the hot side before switching to lower-conductance devices for heat rejection. All I can say is that my thermal philosophy wasn't as robust back then, and I also had an irrational fear of heatpipes :)

I would also wrap the whole TEC stack with Aerogel so heat doesn't leak around the edges. Heatpipes are amazing.

The awkward thing about TEC or Peltiers is that you still need a system to move the heat away from near the device, which often defeats the point.

I was looking into using them to cool the stepper motors on my enclosed 3d printer, but realized that I'd still need a water loop to carry the extra heat away, in which case I could've just setup a direct water loop without TEC. In the end I opted for just heatsinks with fans pointed at them.

Might want to put a TEC on the server CPU!

They were being used for CPU cooling by enthusiasts in the early 2000s, in combination with water cooling. Having a peltier device in it adds so much heat to the overall system that it's really only a measure of last resort in terms of bulk heat transfer.

You also have to be careful that the cold side doesn't get so cold that water condenses around it and into the CPU socket. That was an expensive mistake.

They are relatively expensive, inefficient, and barely work for cooling.

Better to get a non-He cryocooler to make your own liquid Nitrogen drip feed.

One shouldn't go halfway if you want to overclock something. =)

TEC's are pretty cheap and can be had for around $2 for a 100 watts of cooling power unit (plenty for a typical refrigerator).

The thing that stops their widespread use is their very low efficiency. You need big noisy fans to get rid of all the waste heat from the hot side, and that drives up weight, price, size, etc. In nearly every application, gas based refrigeration systems win out overall.

TECs have a very low delta-T limit, rapidly lose efficiency, and those internal losses quickly constrain multistage systems that try to workaround the limits.

Many consider them a niche part with narrow applications. =)

You can still buy Fluorinert on eBay. Get it while it's (not) hot...

Of all the power you pump into one of those modules get 5% cooling and 95% heating. It's practically like running resistive heating.

They're not so much useful for cooling as they are for achieving sub-ambient temperatures where power use isn't a problem. Extremely light and compact fridges/freezers. Final stages of quantum computers to really get down to that zero kelvin.

Can you conceivably use something like this in places where you can't install air-condition (or air-to-air heat pumps)?

E.g., most apartment buildings in Europe, where our windows are not even designed to be used with even portable ACs

Are there any promising thermoelectric materials that might replace bismuth telluride? It seems like everything else is only practical on paper…

TEC's are no good. The best bet is to bury copper tubing deep underground and have 2 loops.

Error 503

science behind thermoelectric cooling

https://web.archive.org/web/20240106121752/https://thermoele...

(With thanks to @varenc)

Reply on news.ycombinator.com