Thermal transistors handle heat with no moving parts | Hacker News Reader

Thermal transistors handle heat with no moving parts

spectrum.ieee.org

"Test Processor With New Thermal Transistors Cools Chip Without Moving Parts" https://www.tomshardware.com/news/test-processor-with-new-th... :

> Compared to normal cooling methods, the experimental transistors were 13 times better.

"Electrically gated molecular thermal switch" (2023) https://www.science.org/doi/10.1126/science.abo4297 :

> Abstract: Controlling heat flow is a key challenge for applications ranging from thermal management in electronics to energy systems, industrial processing, and thermal therapy. However, progress has generally been limited by slow response times and low tunability in thermal conductance. In this work, we demonstrate an electronically gated solid-state thermal switch using self-assembled molecular junctions to achieve excellent performance at room temperature. In this three-terminal device, heat flow is continuously and reversibly modulated by an electric field through carefully controlled chemical bonding and charge distributions within the molecular interface. The devices have ultrahigh switching speeds above 1 megahertz, have on/off ratios in thermal conductance greater than 1300%, and can be switched more than 1 million times. We anticipate that these advances will generate opportunities in molecular engineering for thermal management systems and thermal circuit design.

>Can switch at 1MHz

>can be switched "more than 1 million times"

Seems like longevity is a potential issue. Definitely could be useful for a few applications (especially temperature control), though I'm not really sure about a pure cooling application.

Being able to switch at 1mhz and having to switch 1 million times in a second are different things. You can want a very fast switch but not to do it often.

That's the only way I can rescue these sentences which I agree are a bit confusing.

It's an odd combination of numbers. So if you use it for one second, it breaks?

Though I guess, 1 trillion switches is more than 1 million times too...

If you operate it constantly at it's max switching frequency, yes.

Existing transistors and physical switches etc. will also have silly short lifetimes if you do that math. In practice why would you be switching so much?

> Existing transistors and physical switches etc. will also have silly short lifetimes if you do that math

Power switching in PWM is often 10s of KHz switching speeds and these are expected to last years so not really true.

I mean silly compared to what they will actually be in practice, not that they would also be on the order of a second.

You don't go ok switch is good for 10M actuations, actuation force is x, so it takes this long to press, times by 10M.. switch will last 17.5 months (or whatever it might come to).

You run into those 10 million actuations really quickly on anything Logitech branded.

Not really. Transistors, especially in power electronics, will almost always be run at basically their max switching frequency continuously for almost the entirety of their power-on time. They certainly will not wear out after 1 second of use at their maximum frequency.

(This kind of thing is extremely common in MEMS, BTW: a lot of cool things you can fabricate mechanically in silicon will do something really cool but for a similarly short time period, and so they never leave the lab. All the actual uses of MEMS have much bigger constraints on what you can actually make)

Would there be any returns to pairing this approach with recent advances in laser cooling (and integrated photonics)?

Title is very unclear, but it looks like they're modulating thermal conductivity with an electric field. It's definitely not a peltier junction as I thought at first.

Seems like this is actually a novel mechanism. Very interesting, I wonder what applications it will find? Cooling of CPUs seems like a bad use case, but I'm sure someone will find a very interesting application

Definitely would be really interesting for spacecraft thermal management. Spacecraft only have radiation as a heat transfer mechanism, and they need to be designed to keep all components within a safe temperature under a range of operating conditions (eg: from receiving no sun on the dark side of a planet to full sun a few minutes later). A satellite in low earth orbit can experience a dozen or more sun/eclipse cycles per day, which makes thermal cycling a major factor in long-term reliability. Moving parts are strongly discouraged for reliability reasons, so thermal design needs to be as passive as possible. Having a solid state switched thermal conductance path would be a huge benefit. Imagine being able to switch a radiator off in eclipse to conserve heat and on in sun to reject it.

I agree. Cooling of CPUs is awful use of this tech. Why would anyone turn down thermal conductivity? Just always be sucking heat out.

--------

The most obvious application would be electrically controlled insulators.

If it's winter or summer, insulate the house.

But if outdoors is 72F (or 21C or another ideal temperature), let the ideal temperature into the house.

Temperature controlled crystal oscillators are another application IMO. Applications where we have a 'target temperature', and easy access to differing temperatures (ex inside the oven vs outside the oven vs the heat from a heater)

Perhaps another application is insulation of batteries. You want good thermal conductivity when the battery is in active use at high power draw to sink heat to the environment, but also you may want to insulate it when it's idle and freezing outside, to reduce how much self-discharge is needed to maintain its minimum temperature.

However, it might be simpler to do this with moving parts or phase-change materials, etc

> Why would anyone turn down thermal conductivity?

Why do you think CPU dies can still only be cooled directly over the parts that are generating heat? If I read the article correctly, this tech isn't just the ability to turn off thermal conductivity that already existed, it's the ability to add thermal conductivity that you couldn't before.

We already have high-quality thermal conductors and its name is copper. And copper runs throughout the inside of all of our chips.

I have severe doubts that any "heat semiconductor" would have the same thermal-conductance as copper (or copper heatpipes, and other such heat-devices we have today).

But copper is conductive, so it will conduct electricity too. This could possibly be placed in areas where copper can't be, maybe even much closer to (or even directly contacting) the actual copper traces in the chip.

Unfortunately I don't know enough about CPU manufacturing to speculate much more than that.

You're asking for a thermal jumper.

I don't know much about the inside of CPUs either. But I know that at the PCB-level, we use these devices to carry heat but not electricity.

https://www.vishay.com/en/product/60157/

Copper is still better at conducting heat. But just a little bit of thermal-jumper to make that small electrical separation is all you need.

Datasheet: https://www.vishay.com/docs/60157/thjp.pdf

In any case, modern material science provides us with plenty of useful insulators, conductors, thermal-conductors (but not electrical), semiconductors, and now semi-heat conductors (erm, thermal semiconductor?)

I don't expect the thermal semiconductor to be replacing any of the other materials we already use. Instead, we will find new discoveries and applications for things I can barely imagine.

If you want to conduct heat but not electricity, there's always diamond.

You're right, I forgot that diamond is very easy to grow on silicon during the CPU manufacturing process.

...this was sarcasm, just in case it wasn't already painfully obvious to the downvoters.

This requires a layer of gold below it, and a layer of graphene above it. If you’re worried about shorting things out, this isn’t your solution.

Oof. Good to know.

modern houses should be well-insulated, using heat exchange ventilators and heat pumps get proportionately more efficient in these conditions. i will be shocked if this is ever viable economically.

One possibility you lose from a well insulated house is the heat collection potential of your roof on a cold day. The roof is receiving close to 1,000w/m2 and that energy is just wasted.

Passive houses use a lot of tricks around the angles light enters a home and skylights etc to collect that energy, but a system that uses little energy and doesn’t have moving parts so it can last the lifetime of the home should allow for more options and thus lower costs.

PS: Solar thermal systems work well and have a fairly short payback period, but involve both moving parts and fluids.

> Cooling of CPUs is awful use of this tech. Why would anyone turn down thermal conductivity?

Well, you could build logic gates out of these switching elements ;)

You could switch the thermal sink between multiple radiators to give each one a lower duty cycle.

Heat naturally flows to the lowest temperature object connected by the lowest thermal-resistance (or highest thermal conductivity).

If you have a radiator at 80C, and a 2nd radiator at 60C and both are connected to HotPlate-X, the heat will naturally find itself going to the 60C, while the 80C will more naturally find its own heat radiating out to the environment quicker.

You don't need a thermal-semiconductor to just have heat flow around. That's just... heat physics.

--------

There seems to be some practical applications for this new "thermal semiconductor". Though the next questions are that of cost, performance, etc. etc.

it (the insulator) seems very small. I think, a lot of chips have multiple temperature sensors across the whole chip to decide when to clock down.

If I have 4 cores, and 1 is going flat out, I could see insulating the unused (or lesser used) cores. The idea being give the hot part the full cooling capacity.

That seems like the obvious idea. But maybe I'm misunderstanding or forgetting something fundamental.

You can use this to equalize temperatures on one chip, using it like a changeable heat-shield, so instead of hotspots, you artifically keep some heat from escaping from colder areas too fast which could prevent stress induced cracking. It's better to keep whole chip at 70*C than a one small area at 70 and other areas at 30. Then when heat load from that one small spot vanishes or is moved to other spot, you can change your heat mask on the fly.

> The idea being give the hot part the full cooling capacity.

Isn't it also to prevent the hot part from heating the other, cooler parts? This could be used to give each part of a CPU its own independent cooling. That'd be pretty cool.

> I wonder what applications it will find?

New analogies for Maxwell's Demon. :p

This out of left-field, are there any military applications in when it comes to vehicles' appearance in thermal imagery? I suppose you wouldn't need something like this for the relatively slow switching of stuff like "which side of the tank should be the hot side".

> are there any military applications in when it comes to vehicles' appearance in thermal imagery?

Yeah, you can now display moving images for thermal imagers, like showing the temperature of terrain behind vehicle while moving around.

It sounds like that depends on having outer tiles with basically no heat capacity so that each cools instantly... Or else something ready to pump the heat away.

Yeah, like a thermal transistor. When temperature transfer is blocked, outer surface of that transistor should pretty fast go to ambient air especially with moving air. Remember that transistor is only several atoms thick and can switch at megahertz speeds.

I would think the ideal use case would be for scientific equipment where a very precise temperature needs to be maintained. Right now you either have a large thermal mass to prevent fluctuations (which also means you can't quickly and precisely vary the temperature), or you just accept the thermal noise and try to compensate for it in the analysis.

A device that makes jesus toast (other images too I guess).

"A UCLA-developed thermal transistor uses an electric field to control heat flow, as shown in this illustration."

Uhhh, that doesn't look like an illustration to me? It looks like AI-generated nonsense that has little to do with anything.

The actual paper (linked in another comment here) does have some real illustrations.

Free research paper:

https://www.researchgate.net/publication/375278656_Electrica...

I'm really excited to see progress being made in this area! I haven't read the paper but looked into this area about 10 years ago, and the two questions that jump to top of mind are: 1) What thermal transfer mechanism being controlled?

2) What is the dynamic range in terms of conductance? (Looks like 13x)

There are a number of different mechanisms for energy transfer at this level, and it seemed like the most straightforward way to build a thermal transistor was to get a material that was an insulator for all but 1 transfer mechanism (usually phonons) and then use an electric field to try to permit/block that. The solutions that did exist back then seemed to be very leaky, with thermal transfer while "on" was only about 3x that of when it was off, and this was for microscopic structures where it wasn't obvious how they would physically scale to useful sizes.

I would _love_ a practical thermal transistor technology for an energy management system in a passive house. Allowing your house to "breathe" at night during the summer without things like loud attic fans or having to manually open windows would be fantastic.

We used to build fanless computers. This tech will be very useful in environments that have wide temperature ranges throughout the year from extreme cold to extreme heat. I just took a red eye flight so my brain is scrambled but this tech could also be useful to help dampen rapid temp changes (thermal shocks).

On the extreme cold side, you can use heaters to warm up your components to boot temperature, however if said component is connected your heatsink you are fighting a losing battle.

Feels like there is some messy reporting by IEEE here. The immediate assumption is that this will be applied to CPUs, but none of the researchers' quotes mention this or even suggest it could be viable. There is one quote mentioning power electronics and use with non-silicon semiconductors which seems like a slam-dunk obvious application but that's brushed over.

This is clever technology, but unless it is a heat pump - and if it is, this article is missing something important - then how would it help with chip cooling? Surely the highest conductivity you can achieve is what matters in practice, not being able to turn it off in a nanosecond?

I think "cooling" is just typical science reporting hype. It's a variable insulator. You could use it for holding stable a slightly elevated temperature, like in an oven controlled oscillator.

ELI5 we want to modulate heat flow in a chip, rather than get heat out as fast as possible. Having planted these heat transistors, why would we not just leave them fully on.

We're not going to use heat to represent an analog signal that we want to amplify, or a logic signal that we want to invert, propagate, latch or multiplex.

So what good is a thermal transistor?

Maybe with careful management of heat flow, we can keep areas of the chip at a very constant temperature. That has benefits because then design techniques can be used that are relatively intolerant to temperature variations? Like you know that certain resistances and capacitances will be rock solid.

My condolences for "thermistor" being taken already, by the way.

It would be very handy technology in thermal batteries and combined industrial processes which make use of waste heat.

Of course, you can use other ways to archive the same, but there is something neat in solid-state zero moving parts approach.

Wouldn't a thermal diode be a lot more useful?

A switch can be turned into a diode with appropriate control circuitry. Even for electricity where diodes exists, active rectification is a thing as in certain situations it can provide higher efficiency or signal accuracy compared to plain silicon diodes.

Pretty sure that would violate the second principle of thermodynamics.

>"With that single-molecule layer, the researchers were able to reach the maximum change in conductivity at a frequency of more than 1 megahertz..."

Let's ask Perplexity.ai the following question:

"In Physics, how is Heat related to Frequency?"

Perplexity.ai: "In physics, heat is related to frequency through the emission of thermal radiation by a body at any temperature, which consists of a wide range of frequencies. According to Planck's law of black-body radiation, the frequency distribution of the emitted radiation shifts to higher frequencies as the temperature of the emitter increases. This means that a hotter object radiates mainly in the long wavelengths of the visible band, and as it is heated further, it begins to emit discernible amounts of higher frequency light, causing it to appear white to the human eye. Additionally, the relationship between temperature and frequency is described by

Wien's displacement law of black-body radiation

which states that the

wavelength is inversely proportional to the temperature.

Therefore, as the temperature increases, the frequency also increases..."

https://en.wikipedia.org/wiki/Wien%27s_displacement_law

https://en.wikipedia.org/wiki/Inverse-square_law

https://en.wikipedia.org/wiki/Law_of_squares

Anyway, a very interesting article!

I'm not sure what your point was, but just to be clear the discussion of black body radiation has nothing to do with the switching frequency discussed in the article.

Are you completely sure about that?

And if so, how exactly (show exact steps of reasoning process) please?

In Physics, if something in the Universe (let's call it thing or phenomenon "A") -- is related to something else in the universe (let's call it thing or phenomenon "B") then if B is also related to thing or phenomenon C, then it's very likely that

A is related to C

as well...

Sort of like in logic, if A implies B, and B implies C... then A implies C.

Well, same thing, but with relations/connections in Physics -- if A is related/connected to B and B is related/connected to C -- then A is related/connected to C and conversely, C is related/connected to A...

So it follows (logically) that if Heat is connected with Frequency (and it is, see my original comment) and Radiation connected with Heat (and it is), and Black Body Radiation is connected to Heat (and it is), then Switching Frequency (as mentioned in the article) must be connected to all of those other things via Frequency...

I feel that people might learn a lot about Physics -- by simply studying the known connections/relations -- between various phenomena in Physics...

Maybe you are right and maybe there is no connection/relation between the two -- but I'll let future scientists make that determination by experimentally completely proving that -- or experimentally completely disproving it...

Until that point in time, I know that I for one keep a completely open mind about it, one way or the other...

Reply on news.ycombinator.com