Thermal junctions: diamond lane for phonons?

[McGizmo]

Why not use diamonds at bonds and junctions? I must profess to being a diamond nut in addition to a titanium nut. I'm not referring to the jewelry with diamonds so much as industrial applications of this carbon form. For abrasives, diamonds really can't be beat! I have files, bits, core saws, saw blades and sanding and grinding disks using diamond grit so I knew that the grit was available. After looking into some type of micro ball additive for Arctic Silver or Alumina epoxy as a means of standing off the Luxeon slug and electrically isolating it, I realized that most of the materials used for this are very poor in terms of thermal conductivity. Someone had posted a while back some internet resources giving the various thermal properties of materials. When I looked at diamond, I went duh! I knew diamond was good but to see gold and silver and copper in the 300 to 400 range, aluminium in the 110 to 150 and then there's diamond at 2500! It doesn't need to be a precious cut, single stone, right? All these other compounds are small particles suspended in an epoxy or grease binder.

OK, I'll cut to the chase. Diamond grit can be had in grits from very coarse down to 100,000 grit! The powder is expensive but not really outrageous when you consider that you really don't need a whole lot. I am still thinking and experimenting with different binders and methods of deployment. You obviously don't want to waste this stuff!

I have seen where some of you like to lap both your sink and the slug of the LED. Well if you use a thermal grease with fine diamond powder, lap in place and let the grit embed itself in the materials.

I must be missing something obvious here because I can't figure out why this hasn't been done before or why we haven't heard of this?

Seems to me that it would be possible to make up a slurry of some goop and diamond powder that would be a super conductor and actually better than the slug or the heatsink itself. Why not a puddle of this stuff with say Arctic Silver epoxy as the binder where you have a nice big glop that contacts the sides of the LED slug as well as covers a larger area on the sink. Seems to me that you have provided a super hiway for more phonons to leave the LED faster and be accepted by the sink.

I'm wondering now if I can't make a Luxeon light with Ti. If it is a composite with some diamond rich epoxy vias and surfaces, maybe I can cheat the thermal abyss; at least for short runs?

OK, I have exposed my ignorance here. Please educate. /ubbthreads/images/graemlins/grin.gif

BTW, for those who are inclined to follow my folly, I found the following Diamond Powder Source . Maybe I can sell the idea to Loc-Tite and call it Diamond Lane Thermal Bond? Don't worry, I won't quit my day job; I don't have one. /ubbthreads/images/graemlins/icon15.gif I think it's called self-deployed?


Don

 

[Minjin]

Don,

Titanium...diamonds...unobtanium lights and LEDs...

You really do have way too much money. /ubbthreads/images/graemlins/smile.gif

I'd go for the Diamond Spray. Hell, you guys talk about HA3 and EN coatings...can you say diamond coating? /ubbthreads/images/graemlins/wink.gif

Mark

 

[Y2Kirk1028]

I found this review article on thermal grease, check out the diamond grease here

 

[yclo]

It's very interesting that you mentioned that, because in another thread a couple weeks ago. There was a link to a table with the thermal conductivities of varies materials and Diamond came up very high.

I think while you're at it, make the body out of silver so conduct even more heat!

 

[Slick]

I like the way you think Don. I'm into pure over-kill as an engineering method too. However, I think diamond-slurry is nearing the equlvilence of killing flies with a sledge hammer.. /ubbthreads/images/graemlins/grin.gif /ubbthreads/images/graemlins/grin.gif /ubbthreads/images/graemlins/grin.gif

I haven't ponied up the cash to move toward such exotic materials because I haven't had any problems with heat related failures direct driving WAY over spec with my die polishing technique. I'm not really driving any of my 5W's past 1.2amp so they're not really that far over.

The truth is, I'm stil salivating over that custom McLux head you made that hosted the 30MM collimator. :P

 

[IsaacHayes]

I think you should team up with that guy that repairs fried luxeons. Have him take the slug off of the die, and add additional bond wires. Then use diamond grease for a 1 thermal joint directly to a heat sink and run the 5w at like 15watts or more! /ubbthreads/images/graemlins/grin.gif

 

[Doug Owen]

You should be aware that there are two types of 'diamond dust' out there. There is the natural (single crystal), typically mined in South Africa, and the man made (poly crystalline) made by the 'du Ponte' (explosion) process. The latter is by far cheaper (about a third the cost) and no doubt totally unsuitable to the task. Unless you specify, I bet you get the poly crystalline.

We use both at work. The natural can be seen as a bunch of random shards in the Electron Microscope, the man made looks like popcorn. The man made one is perhaps the better abrasive, especially at the sub micron level we work in, but is sure to be a poor conductor of heat by comparison.

There are also some interesting forms of graphite that have diamond like heat conduction but also can be electrically conductive in a different direction. No practical use here that I can see. Graphite is sort of a 'two dimensional diamond' if you will, bonding in sheets rather than cubes.

The important aspect is most any good 'heat sink compound' will perform well if set up properly, none will fix poor workmanship. The idea is the gap is very small (or else....). The thermal conductivity of the rest of the path while far better than the thin layer is also much longer. Given reasonable care, it will dominate the equation. That is even a perfect thermal compound will not provide much better cooling if the layer is thin. We're not living on a degree or two better cooling here, right?

Doug Owen

 

[McGizmo]

[ QUOTE ]
Doug Owen said:
......... The idea is the gap is very small (or else....). The thermal conductivity of the rest of the path while far better than the thin layer is also much longer. Given reasonable care, it will dominate the equation.........
Doug Owen

[/ QUOTE ]

Understood but if the thermal compound is superior in thermal conductiviy, then it needen't necessarily be thin, right? I agree that I haven't had any failures using conventional methods of thermal bonds. I have noticed the luminous output drop due to what I expect to be elevated temps of the LED. If current methods are not a bottle neck in the migration of phonons, then there is no point in improving on the junction. However, if they are a bottle neck or if you can increase the size of the path by the additional contact of a good thermal material contacting the sides of the die slug as well as spreading out and making more surface contact with the sink.. I guess what I'm getting at without stating clearly is the notion of a diamond rich compound itself becoming the primary heat sink. This seems to have merit if a diamond rich compound can have a in place thermal conduction that is significantly better than the materials it joins. Maybe it is a case of overkill or splitting hairs. I have held various lights up to a light meter and watched the luminous output start dropping almost immediately. Any improvement on phonon passage or capacity would slow this process, me thinks.

I think I'll have to make a carbon fiber flashlight vacuum baging diamond rich epoxy into the cloth. /ubbthreads/images/graemlins/grin.gif I can throw some coarse grit on the surface and have a good non skid surface. Then I can take my plastic flashlight and cut a SF flashlight in half with it. /ubbthreads/images/graemlins/grin.gif

- Don

 

[Doug Owen]

Don,

I think you need to rethink the problem. First off, there really is no such thing as a true heatsink, something that will eat up all the heat you give it and not get hotter. Materials have what's called 'specific heat', how much energy you have to put in to raise their temperature one degree. But if they can't pass the energy along to the next element, they will get hotter. No 'magic bullets' here.

Yes, a two fold improvement would allow you to double the thickness, but that's not the key to the castle either.

FWIW heat doesn't move in photons, there is no quanta for heat (unless we're talking IR), photons are the quanta of Electro-magnetic radiation, not heat. Heat moves by three means, conduction, convection and radiation. At the temperatures we're using only the first two, with conduction being by far dominant. Convection plays a role in getting the heat out of the flashlight body of course, but a minor one I suspect. We need to focus on conduction.

You can also measure the actual LED temperature if you've a mind to. Notice the -2mV per degree C change in Vf? Just measure it (quickly) at room temperature, hold the current constant, and do the math. From my modest fiddling I think most of the light output drop I've seen is a result of lower drive current, however.

Remember, there's a 15 degrees C per watt factor from die to slug we can't do a thing about. Shaving a few tenths of a degree C per Watt (or even a few whole degrees) off the slug to heatsink junction is getting deep into 'diminishing returns' country.

I do, however, like the diamond impregnated flashlight body idea. Just the ticket for 'flashlight wars'......

Doug Owen

 

[McGizmo]

Doug,

Thanks for the info and explainations. I know we can't do anything about the die to slug connection and if that is the weakest link in the thermal path of a particular design, then there is no reason to go further; as i understand it. However, in cases where the slug needs to be electrically isolated from the sink such as red LED's and Wayne's wizard drivers, I think I'll try floating the slug in some diamond rich goo. I prefer this idea at least conceptually to using two coats of Arctic epoxy or some other electrically isolating layer further on.

- Don

 

[tvodrd]

I read somewhere that the thermal conductivity of diamond films is about as awsome as it gets. Could provide a means for Lumileds to better manage the junctions. I like the idea of sub-micron diamond dust heatsink compound. Is there an "Arctic diamond" paste on the horizon? /ubbthreads/images/graemlins/grin.gif

http://www.stanford.edu/group/microheat/two.html

Larry

 

[McGizmo]

Larry,

Interesting article there! Not sure I can understand it or more important use it. /ubbthreads/images/graemlins/grin.gif I do have some 100,000 (1/4 micron) grit diamond powder coming And I am assuming if I use a bunch of this with a little of 220 grit powder, I can stand the slug off for electrical isolation and yet provide a reasonable thermal path.

I also noticed that silicone has a reasonably decent thermal conductivity. Do you suppose that this would apply to "pure" silicone grease as well? I am wondering if silicone grease should or could be used as the binder for a diamond grease?

Wayne, Leo and I have discussed a thin PCB with a hole in the center as a "generic" guide or mount for Luxeons. Such a board could be relatively easy to center on a sink and glued in place. The hole in the center would serve as centering locator for the Luxeon and the soldered leads would hold the Luxeon in place. This would allow for a thermal grease to be used in the junction bond instead of an epoxy. I like the idea of dropping the LED into a diamond pot. /ubbthreads/images/graemlins/smile.gif (I can see Doug shaking his head in disgust)

If anyone has a take on the silicone grease, please pip up.

- Don

 

[Doug Owen]

[ QUOTE ]
McGizmo said:

If anyone has a take on the silicone grease, please pip up.

- Don

[/ QUOTE ]

(Assuming that's "pipe up", not "'pip up" (which I've no idea how to do), sure, works great. Used it for the last 30 years or so. Messy stuff, *never washes off* in 'anything'. Deceptively heavy.

There's a 'funny story' from the early days of Seagate (the computer disk drive folks). Back in the dark ages when HDs were a serious challenge they built a brand new building to control contamination. Principle amongst the 'can't seem to get rid of' was the 'goose grease' (Silicon HS compound) used in the electronics where they mounted to the drive case. A line was painted across the floor, no drive tools, fixtures etc. were allowed to cross. The line moved 3 times before the building was abandoned.

Anyway, Wakefield and the heat sink makers sell goose grease. As I assume does Radio Shack.

Doug Owen

 

[McGizmo]

Thanks Doug. If my spell check (try to use it if time permits) accepts a word, I'm doomed. So you never piped before? Don't know what you're missing. I have all kinds of silicone grease (clear stuff) used for UW camera gear. I also have some GLY Silicone grease that is supposedly pure silicone grease, what ever that means......

- Don

 

[tvodrd]

Don,

In my limited experience, silicones, particularly the rubbers have poor thermal conductivity. The guys at G.E. silicones tried to help me with an elastomeric heatsealing app by adding as much magnesium oxide(?) as they could get into it. Wasn't good enough. I suspect the fluids/greases aren't much better. They do offer extremely low vapor pressures- as in don't dry-up. The low viscosity stuff we use at work does migrate like hell. I lucked-out and was able to put that property to use /ubbthreads/images/graemlins/grin.gif. There's a new thermal pad for CPU heatsinks that has a lot of people raving. (Back with a link, I hope)
http://www.overclockers.com/articles742/ I've seen a couple other articles that concluded this stuff is better than AS. Shin Etsu's site has the stats on it.

Larry

 

[Icebreak]

Don –

I’ve been meaning to get a definition of phonons that would be understandable to a severely knowledge challenged science fan like myself. I found this:

“The evidence on the behavior of vibrational energy in periodic solids is that the collective vibrational modes can accept energy only in discrete amounts, and these quanta of energy have been labeled "phonons". Like the photons of electromagnetic energy, they obey Bose-Einstein statistics.”

here:

http://hyperphysics.phy-astr.gsu.edu/hbase/solids/phonon.html

Interesting page.

I posted this because it helps me to understand what you guys are talking about and I thought I might not be the only one that didn’t know what a phonon was.

On your proposal for a diamond slurry, I’m sure not the guy that would know, but my instinct tells me this has a good possibility of being one of those ideas that is so simple in concept that folks may say, “Now, why didn’t I think of that?” The only caveat I see is that the frequency of the media and the frequency of the grit need to have some harmonics, otherwise the pieces of grit accept heat well but don't transfer the heat through the media (with good efficiency) to the next piece of grit.

Your diamond slurry concept conjures up another idea that has been discussed here before. Since Cree subtracted sapphire (Al2O3) from the process of building LEDs I'm still hoping someone will use this material in building a flashlight. A diamond slurry may allow this material and others to be stacked to make a super slug. Sapphire to silver to aluminum to copper. Of course, this may be why Icebreak doesn't need to be designing heat sinks...he would probably end up building a battery and claim he had a magical flashlight. /ubbthreads/images/graemlins/wink.gif

- Jeff

 

[PsycoBob[Q2]]

From what I remember, the Sapphire chip in LED's is used primarily as electrical isolation... can it be replaced with a CVD diamond layer? Likely not cheaply, but it may allow for a 15w+ LS with that change, and a copper slug. Does anyone know what the thermal resistances layers of the doped silicon/sapphire/Al slug assembly?

If the diamond can be deposited to the WHOLE Cu slug, to prevent corrosion and to allow electrical isolation, then the LED chip is applied to that... Interesting, but Lumileds would likely slap me and point out a few hundred reasons why not.

 

[Rothrandir]

what about a silver slug? /ubbthreads/images/graemlins/grin.gif

 

[Doug Owen]

[ QUOTE ]
McGizmo said:
Thanks Doug. If my spell check (try to use it if time permits) accepts a word, I'm doomed. So you never piped before? Don't know what you're missing. I have all kinds of silicone grease (clear stuff) used for UW camera gear. I also have some GLY Silicone grease that is supposedly pure silicone grease, what ever that means......

- Don

[/ QUOTE ]

Sorry, Don, wrong stuff. Silicon not silicone....

Tell ya what, we're both in Berkeley, unblock PM and lets exchange contact info. I'll give you some goose grease to fiddle with maybe show you some seriously fun Big Science, set yer chickens free.....

BTW, I know what you mean WRT spell check. You should see some of the stuff I get into.....

Doug Owen

 

[McGizmo]

Doug,

Thanks for the offer. I haven't enabled PM because it doesn't work for me and I guess a mesage gets bounced back to the server where it sits, clogging the arteries. If you are so inclined, please feel free to e-mail me. If not, no problem.

Thanks again.
Don