backtop


Print 36 comment(s) - last by NicePants42.. on Jan 25 at 2:52 PM


Professor Peidong Yang led the University of California, Berkley team who discovered silicon nanowire's thermoelectric properties.  (Source: University of California, Berkley)

A scanning electron microscope image of a thin silicon nanowire stretching between two heating pads, one as a heat source, the other as a sensor.  (Source: A. Hochbaum)
Silicon heat capture could allow cheap refrigeration and energy saving

Among the many valuable properties of silicon is its ability to capture solar energy to create electricity via its photoelectric character.  Now scientists are discovering that silicon, when properly prepared, can form a very good thermoelectric as well.  This opens the door to a plethora of uses, including refrigeration, solar heat power generation, and power generation from other heat sources, such as computer waste heat or car heat.

A thermoelectric device has two  basic modes of operation.  When a thermoelectric is placed over a heat gradient, it generates an electric current.  The other mode is the reverse; when a thermoelectric is exposed to an electric current, it creates a heat difference, cooling one side of it, and warming the other side.  Thus thermoelectrics are applicable to power generation, refrigeration and heating.

Traditional thermoelectrics, which have been around since the 1960s, rely on either bismuth telluride or lead telluride.  These materials are relatively expensive due to scarcity and lack of a large manufacturing infrastructure.  They are also bulky and require more material, which further increaese their cost.  While thermoelectric coolers have achieved modest commercial usage in seat coolers and picnic coolers, they have yet to realize their full potential.

A new breakthrough may change that.  Professor Peidong Yang and his colleagues at the University of California, Berkley published in last week's Nature journal the results of years of research into using silicon as thermoelectrics.  Their results show that silicon can be a viable thermoelectric.

The key is in the preparation.  The researchers prepared thin nanowires of silicon.  When these wires are exposed to a temperature difference, they generate electricity.  Standard silicon is a poor thermoelectric, but according to Dr. Yang, "the performance of the nanowires is already comparable to the best existing thermoelectric material."

A good thermoelectric needs to have two key properties -- good electrical conduction, and poor heat conduction.  Silicon typically conducts both very well, but by producing 50 nm nanowires, the heat conduction of silicon is reduced to one hundreth of its normal levels, while electrical conduction remains unchanged.  The material is comparable to commercial thermoelectrics.

Two possible uses of the technology are to generate electricity from waste heat of car engines.  Current thermoelectrics are too expensive and large to make this a practical possibility.  Nanowire silicon layers, though could provide a means to recapture some of the energy lost to heat during the conversion to mechanical energy in a car engine.  This extra savings could be stored in batteries, to give next generation electric vehicles, such as the Chevrolet Volt, even better efficiency.

It could also find a home in solar power cells.  By coupling it with traditional photoelectric cells, much higher efficiencies could possibly be reached.  Yet another application is to put the materials in computers to provide energy savings, which would be particularly valuable to mobile computing.  Further, it could be used in refrigeration applications, as well.

Much work needs to be done before the process is perfected.  The physics behind why nanowires of silicon lose their heat conduction is not understood, which stands in the way of refining the efficiency of this class of devices.  Further creating a thermoelectric on the macroscopic scale, by creating a network of nanowires, has yet to be accomplished.  Still, the discovery of these properties in silicon promise a way to eventually use replace current less ideal thermoelectrics with an abundant material with a large processing infrastructure.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

RE: Not that universal
By MrTeal on 1/21/2008 6:32:34 PM , Rating: 3
quote:
Let's imagine a TEC that conducts NO heat. If we were to stick this TEC between a CPU and a heatsink, would heat fail to flow from the 'hot' side to the 'cold' side, and thus melt our CPU?


It probably wouldn't melt the CPU, just a small part of the die. ;) n=1-Tc/Th. If the ambient temperature was 0, then a perfect TEC might be able to convert all the heat into electrical energy. With the cool side at room temperature, even a perfect TEC would not be 100% efficient, and heat would build up until the device was destroyed.

In real world applications, if you have a device that's converting a set amount of power into heat, it will continue to get hotter until the temperature differential is such that an equilibrium is reached. Some small portion of the power will be converted to electricity and moved elsewhere, while the majority will be transfered to the heatsink and then to the ambient air. The only difference is that now the increased thermal resistance will mean that the temperature difference from hot to cold is greater than without the device.

Basically, this kind of device might be good at generating some power where heatsinking isn't currently necessary. I used to have an old Toyota Tercel, and in the winter I'd stick a piece of cardboard in between the grill and radiator. If I didn't, the coolant never got hot enough to properly heat the inside of the car when it was -30 outside. It was easier than changing the thermostat. In that kind of case, this would make sense. Large delta-T, not worried about overheating the engine.

Now, picture someone coating their radiator in the summertime in LA. Already the cooling system is working as hard as it can to keep the car from overheating in the 100F+ weather while you're stuck in traffic. If you place a layer that decreases your thermal conductivity you either have to increase the surface area of your radiator, increase the conductivity in other parts of the system, or build the engine to survive higher temperatures.

In the end, these would be useful if you have heat that's just wastefully floating away. Basically, all I'm saying in the OP is that if you've invested energy (running a fan) and materials (coolant, radiator, heatsink, etc) towards maintaining a low temperature so that your product doesn't fail, chances are trying to use thermocouples to retrieve a little excess energy will not be practical. Efficiencies would have to get well above the current TECs, which are the same as what these prototype have, before many new applications could be found. They work great in things like radioisotope generators, because you can have a hunk of plutonium releasing 4500W of heat and glowing red to produce 300W of electrical power.


RE: Not that universal
By DKWinsor on 1/22/2008 12:33:09 AM , Rating: 2
I think in my last post I mixed up hot and cold side, I apologize if this confused anyone. In a TEC running in peltier mode the cold plate gets cold and the hot plate gets hot. You want to cool the CPU not heat it, so you put the TEC's cold plate next to the CPU. This is different from running a TEC in the seebeck mode, where the orinentation of the hot and cold plates do not matter much - the plate on the side of the CPU will be hotter than the plate on the side of the heatsink. All that changes when you flip plates is you flip the voltage to negative, which is easily changed by reversing your wires. I hope that's clearer than mud.

That's funny, because my dad did something to his Toyota Tercel's manifold with silicone sealant and I think cardboard.

But sticking a piece of cardboard between the radiator and the windflow is not the same as sticking a TEC between a CPU and a heatsink. Sure they both increase the thermal resistance but one has a special thermoelectrical effect that counters it.

I've used a TEC to cool my CPU - I put it in, hooked it up to a water cooler, insulated it so water wouldn't condense (the CPU went below ambient), and ran it on my 12v line. Since it was rated at 14.4v it moved less heat than the maximum it could, but the heat it did move was moved more efficiently per watt, with less waste heat output to the waterblock. I no longer use it, just use water, because it was a hassle and ate a lot of power. Plus that was on my old CPU and my new Core 2 Duo overclocks fine as is. So anyways I have experience using the peltier effect of a TEC but not the seebeck effect.

Let's suppose we have a CPU that is cooled by water, so a high wattage of heat dissipated by the CPU equals a low temperature increase in the water. But instead of some thermal compound, let's stick a piece of cardboard between the CPU and the waterblock. Our CPU when turned off is at ambient. We turn it on, and within a second the CPU temp rises by some wattage, which equals 5 degrees. Now there is a differential of 5 degrees between it and the water. A differential of 5 degrees means 1 degree's worth of heat wattage can leak past the cardboard and into the water, which for the purposes of this remains the same. The CPU has been coolde by 1 degree so it is 4 degrees above ambient. In the next second that the CPU is on it again rises by 5 degrees worth of heat wattage. This time the differential is 9, and therefore 1.8 degrees can leak over. It's at 7.2, we add 5, and a 12.2 differential means we dissipate 2.44 degrees. And so on until the CPU is 25 degrees hotter than the ambient water temperature and now the CPU dissipates as much heat into the water as it generates.

Let's do the same thing with a regular silicone nanowire TEC with poor heat conductivity. The CPU turns on and it raises 5 degrees. 1 degree of heat on the CPU side leaks out so it's again at +4. Now here comes the tricky part - where it leaks out to - because I don't understand the seebeck effect. One of 2 things could happen.

The first would be this that 1/2 degree goes into the water and 1/2 degree of heat is turned into electricity. If so, then our TEC is as thermally bad as cardboard, allowing the CPU to reach +25, but it also decreases the water's temperature and generates electricity, both of which are good.

Or the second would be that that full 1 degree goes into the water, and we still have to take care of the seebeck effect. I don't know if the water side gets hot or not by this effect, but I do know there is a current generated. Due to conservation of energy this must come from the hot CPU side. Let's say it generates as much power as needed to further cool the CPU by 1 degree so it's at +3. In the next second the CPU is at 8 degrees, so 1.6 degrees go into the water and 1.6 is converted into electricity. Continuing this on, you find the CPU is only +12.5, and you're generating power.

Either way, if you can afford for your CPU's heat to increase over what it is now (and I have no idea by how much), then go ahead and put a thermally resistant thing in your heat loop because you'll get back energy with less waste heat going into ambient.


RE: Not that universal
By MrTeal on 1/22/2008 1:25:47 AM , Rating: 3
quote:
Let's suppose we have a CPU that is cooled by water, so a high wattage of heat dissipated by the CPU equals a low temperature increase in the water. But instead of some thermal compound, let's stick a piece of cardboard between the CPU and the waterblock. Our CPU when turned off is at ambient. We turn it on, and within a second the CPU temp rises by some wattage, which equals 5 degrees. Now there is a differential of 5 degrees between it and the water. A differential of 5 degrees means 1 degree's worth of heat wattage can leak past the cardboard and into the water, which for the purposes of this remains the same. The CPU has been coolde by 1 degree so it is 4 degrees above ambient. In the next second that the CPU is on it again rises by 5 degrees worth of heat wattage. This time the differential is 9, and therefore 1.8 degrees can leak over. It's at 7.2, we add 5, and a 12.2 differential means we dissipate 2.44 degrees. And so on until the CPU is 25 degrees hotter than the ambient water temperature and now the CPU dissipates as much heat into the water as it generates.


Ok, I think here's where you're getting it confused a little. In the simple cardboard example you have, the cardboard will have a thermal conductivity, k (kappa), given in W/(m*K). The thermal resistance (theta) is = length/(k*Area). So, in this case, say the cardboard is 1mm thick (l) and has a 4cm^2 area. The whole thing has a thermal resistance of 10K/W, so for every watt of power it dissipates, the temperature will rise 10 Kelvin (or degrees C).

There will be some delay involved due to the (small) thermal mass of the CPU die, but assuming that the CPU die has perfect contact with the top of the package, and the waterblock stays at ambient the whole time (ie, it has no thermal resistance), the temperature of the CPU die will be...
Tcpu = 20C + 10(K/W)*P, where P = power dissipated and T is in celcius. So, if your CPU is dissipating 40 watts, with a crappy piece of cardboard in between it and the waterblock it'll eventually reach an equilibrium, at 420C. Ouch.

For reference here, when you're insulating a transistor in a power amplifier or some such thing, you often have to electrically isolate the back of the transistor from the heatsink, since the transistor might be hooked to a voltage and you don't want to zap someone. One of the best materials was beryllium oxide, until they stopped using it on account of the cancer and all. It had a thermal conductivity of about 0.25K/W. With just one of those crappy elastomer pads you see on heatsinks, it's about 0.1K/W. Hence why people buy the fancy thermal grease.

quote:
Let's do the same thing with a regular silicone nanowire TEC with poor heat conductivity. The CPU turns on and it raises 5 degrees. 1 degree of heat on the CPU side leaks out so it's again at +4. Now here comes the tricky part - where it leaks out to - because I don't understand the seebeck effect. One of 2 things could happen.


If you think of it as a heat engine, you have power going in from the CPU, say 100W. You have electrical power out, and being exceedingly generous I'll say 10%, so 10W. The other 90W are heat that goes to the water block. You would still be dissipating 90W, and even if the element had a resistance of 1K/W, you'd still be looking at a 100C cpu once it stabilized. All that for 10 lousy watts of power.


"Intel is investing heavily (think gazillions of dollars and bazillions of engineering man hours) in resources to create an Intel host controllers spec in order to speed time to market of the USB 3.0 technology." -- Intel blogger Nick Knupffer

Related Articles













botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki