Anodizing aluminum to mount bare emitters... why?

Jayls5

Newly Enlightened
Joined
Oct 11, 2009
Messages
74
I have seen this done in order to get the bare emitter on solid aluminum. Since it's so difficult to get the solder to attach, I have seen people anodize the area and it accepts reflow solder better. However, anodizing.... or adding a layer of aluminum oxide to the surface, lowers the thermal conductivity from 237 W·m−1​·K−1 ​to 30 W·m−1​·K−1​. This is bad... like nearly as bad is Titanium bad (which is rated at 21.9).

If the whole point of direct mounting a bare emitter to a heatsink is to bypass the inefficiency of the star, why would anyone do this? Are my numbers wrong?
 
How many meters thick is anodization? Plating is another option I've seen.

I don't know of a way to accurately measure the difference between well-applied thermal adhesive and well-applied thermal soldering.
 
The thickness of the anodization doesn't have much of an effect on thermal conductance. Ironically, thickness does have an effect on emittance. Anodizing has a dramatic positive effect on aluminum's emissivity.

Your conductivity numbers aren't wrong, but they don't really translate exactly to the specific situation. The ability of alumina to conduct heat is only one part of a 3-dimensional puzzle. The first thing that must be considered is that heat conducts through matter in two fashions- in-plane and through-plane. It's pretty much as it sounds- in-plane is how it moves through a single mass and through-plane is how it moves between two pieces of material. As I have written about before, you could take two pieces of the identical material and mate them together as closely as possible and they won't conduct the same as the single solid mass.

Heat is the energy of excited electrons. Fourier's law states that heat moves towards cold. In a sense, cold doesn't really exist. Only the lack of heat. If you think of this on the atomic level, the energy of one excited molecule is easily transmitted to the one atomically bonded to it. Think of them as a bunch of spinning tops- as one bumps into the one next to it, it transfers energy. Once it comes to the end of the mass, it can only transmit energy at the rate the material contacting the mass is able to accept. If you do this in a vacuum, the energy will stay in the mass- it has nowhere to go. If you do this in air or water, heat moves at that material's ability to conduct heat. (We're leaving out emissive radiation but that's another story)

This is where conductivity becomes less absolute. Each excited molecule has a certain amount of kinetic energy. When molecularly bonded with another molecule it transfers energy quite well. But that kinetic energy also allows the molecule to transfer energy to another molecule that's very close to it but not atomically bonded. A simplification would be that the energy is able to jump the gap.

In a practical application, the closer these two surfaces are to each other, the higher the level kinetic conductance. As you move these surfaces away from each other, this effect diminishes logarithmically. At the atomic level, the smoothest, flattest surface finish we are presently capable of achieving looks like a field of boulders. Actual molecule-to-molecule contact is a very tiny fraction of a percentage point of total "contact" area. Everything else in between has some other element filling the voids. If you were able to place molecules the same as the mass in those voids, they would atomically bond. So, the closer you can get two surfaces together, the more kinetic conductance you can achieve. This is the reason why the thermal conductance rate of alumina (anodization) isn't absolute in a practical application. Now, if you build a significant layer of alumina on the surface, eventually the kinetic effect is lost and now you are dealing with a material that conducts at a much lower rate.

This kinetic effect is the reason all the manufacturers of thermal products direct the thinnest application as possible. I have seen a few articles where computer cpu coolers were evaluated. In those articles, a rough contact surface was smoothed and cooler efficiency improved.

Plating accomplishes not a whole lot. If you plate a piece of copper with silver, what you have is one very conductive material transferring heat into another, more conductive material very efficiently- but only for a very brief distance. It's like driving to Grandma's and beginning your trip by driving to the end of your driveway at 250 mph. If the rest of your trip is at 35, you accomplished little. The other factor with plating is that it's a chemical bond, so while the conductance is more efficient because of the closer relative distance of molecules, it's not absolute as if identical elements were molecularly bonded. Plating is often used to prevent the oxidization of the base material. As most metals are reactive with oxygen, as soon as their molecules are exposed to oxygen they begin to corrode. But silver doesn't really oxidize any much more differently than copper so there's not much to be gained. If you are plating for the purpose of allowing a diode to be soldered to aluminum, that's alot of effort for not alot of gain.

To return to the practical application of the OP, anodizing aluminum to allow the soldering of a diode is not an efficient approach. Unless a portion of the overall physical design dictates this approach, there are better ways that will conduct heat more efficiently.

Keep in mind, conductivity is a function of how well the material conducts and how much there is of it. Think of it like plumbing- bigger pipe/higher pressure is going to give more output that smaller pipe/lower pressure. You can balance the two in order to achieve your desired effect.

In the application of conducting heat from a diode, you have to consider each and every element in the chain of process. The first thing the diode contacts is the bonding agent. If you glue it, there are bonding agents that conduct between a fraction of a point and (reportedly) up to 60 W/ mK. Solders can range from the mid 20's up into the 70's. If you affixed your diode with 50/50 solder, none of what comes after it matters all that much. Once you're past that very tiny bottleneck of the led's thermal pad, it becomes simpler. If the design that required a very small footprint had significant thermal output, the builder would be much better served to mounting the diode to a copper post and affix it to whatever the base material might be.

I saw a build where the diode was affixed to the top of a small-diameter copper post. Only the thermal pad in the center of the diode was in contact with the sink. The electrical contact pads were over voids which allowed wires to run underneath the diode. This gave away all the thermal transfer ability of those two pads- a significant amount of area. At the same time, this design was conducting heat out and into the wiring. As I'm sure you know, the hotter the material, the more electrically resistive it becomes.

Thermal management requires a holistic approach. You must understand every facet of the design's thermal flow and focus on the bottleneck. Nothing in the design will conduct any more effectively than that point.
 
Thanks CTS, that was a nice cohesive explanation. I knew there was something wrong when I saw people doing this. I already view this in terms of bottlenecks, and I saw this practice as exchanging one inefficiency for another.
Just from a logic standpoint, is sounds like the best route is press thermal epoxying a bare emitter directly to heatsink with something like arctic alumina or silver adhesive. The epoxy can be electrically non-conductive, so you still get the benefit of all of the led pads contacting the through the thin layer to the main body. The disadvantage is the difficulty of soldering wires to the top of the board to power it. It simply sounds like an easier and more efficient contact patch than direct soldering to a poorly conducting anodized layer of aluminum.
Obviously I am leaving emissivity out of the equation here. I see emissivity playing a larger role after a heatsink has become fully saturated. This seems more important on a light not being held in someone's hand, which is essentially a giant 98.6 degree stable convection pump. Does everything I am saying sound about right to you? I am in the design stages of making an awesome mini light, and I am hoping to avoid active cooling to improve run-time.
 
Pretty much. I would suggest you re-examine your diode mounting strategy though.

First off, AS5 conducts at about 10 (maybe) where the worst of solders (50/50) conduct at 20. 97/3 conducts in the low 60's and indium in the 70's.

But let's not put the cart before the horse.

You need to look at your thermal path as a part of your overall design. And in doing that, lets use a well-known build as our example. The Jayrob Maglite XML conversions is an excellent example of a useful, common-sense build. He's driving the XML to 3 amps, uses a very small aluminum MCPCB and mounts to a moderately sized heat sink that's slip-fit into the anodized ID of the maglite. To my knowledge these lights have no issue with overheating or reduced outputs due to poor thermal management. If he were running 4.5A, that would probably change. My point is this- are you generating so much heat with your design that you can't get it out of the diode fast enough? If so, why are you using that diode? If you want to be the guy that can say "I drove an XML at 7 amps", then OK. But maybe that diode is a poor choice. Part of that choice is the PRACTICAL limitations of each element of the overall design.

The diode is only going to be able to conduct so much thermal energy out of the bottom of the die. Unless you somehow have a way of furnace brazing that die to a copper plate prior to its assembly, about the best you can do is get a bond at around 70 W/m K. The smaller the die, the smaller the amount of energy flow. The warmer the diode, the less light it emits and the more input energy is turned into heat. You will eventually reach the point where the diode won't go any further. But getting there, the additional energy is simply going to waste. That's why when you put a 1000 lumen and an 1100 lumen flashlight side by side, you're hard pressed to identify which is which.

Back to the Jayrob Mag. At 3 amps he's having no problem with a very small diameter board (12mm?) made of a pedestrian material, aluminum. If he wanted to push to 3.5, he might have to address a larger heat issue. One of the limitations of his design is getting the diode up into the reflector. So if he wanted to go to 4A, maybe the approach would be to ditch the Mag reflector and use one that can be mounted right to the board. On the other hand, he could use a more thermally efficient board.

One of the most significant limitations of modern MCPCB's is how they supply current to the SMD-style LED. This requires a barrier mask. If you sectioned a board, you would find the base material, then an electrically-insulating film. Above that you have very thin layer of copper, then the top mask. That thin layer of copper is isolated into three sections, the positive side that opens above the top mask to the + side of the diode, the negative side that does the same and then the thermal pad in the middle. This typically covers a center bar and two D-shaped parts on each end. What this is doing is taking heat that flowed through solder that conducts at 60, into copper that conducts heat at 400 and the copper then disperses that heat into a large area which conducts into an electrically-insulating film that conducts heat at about 1.5. Thermal conductance is two elements, speed and volume. High conductivity through a small area is just about as useful as slow conductivity through a large area. Your design dictates the best approach.

To push the Jayrob light to higher outputs, you could utilize a copper-cored MCPCB. They are very similar in construction to the board described above, except for the thermal pad. In this design, the pad under the LED is raised slightly. This projects up and through the dielectric mask, allowing the diode to directly attach to the board's core. This is a much more efficient way of conducting heat than direct-bonding the diode's thermal pad to the heat sink while not sinking the connection pads. If you really wanted to get fancy, you could flow the LED onto the top of the board using 97/3 and flow the board to the sink using a lower temp solder like indium. But you'd need to do the math to understand if it was necessary or useful.

I'm working on a design right now that goes the other direction. I'm using a 35mm board and bonding a reflector centering ring directly onto the top of the board. The reflector can now get the diode up in where it needs to be with a very small opening. I'm using a very inexpensive and low-tech approach to managing excess heat as my design has the space to accommodate it- I'm using a 42mm reflector on a single XML. A triple XML in a P60 format would present a challenge.

One more consideration is thermal equilibrium. If you heat one portion of a mass to a certain point, then stop, that heat will continue to move through that mass until each particle is at the same temperature. In effect, if you heat one end of a foot-long copper rod from 50 degrees to 200 degrees, the cold end will increase in heat while the hot end will decrease. At some point, that temperature will stabilize, dependent upon at what rate the energy is being conducted from or emitted by that mass. This is important to understand in your overall design. If you move the heat efficiently from the diode to the board to the sink and into the flashlight's body, but fail to provide a path out of that mass, eventually the assembly's mass will saturate and conductivity will slow and eventually cease. This is the point where energy input must be reduced to the point that maintains the device's ambient operating temperature. If you have four pounds of copper and one 18650, there's nowhere near enough energy to approach thermal saturation before you run out of juice in the battery.

Your design should entail each of these considerations. Why use an expensive component or material or employ a difficult to execute construction aspect unless it's necessary?
 
If the whole point of direct mounting a bare emitter to a heatsink is to bypass the inefficiency of the star, why would anyone do this? Are my numbers wrong?

I'll explain what I do and why. I make dive lights for buddies using the SST-90. I machine a button of sorts that the bare LED mounts to on one side and the driver for it on the other. The end with the diode screws into the reflector. Then that assembly is pressed into a waterproof aluminum housing I machine.

I anodyse the aluminum for one reason, electrical insulation. I have holes drilled from top of the button to the bottom where the driver is and wires pass through there which carry a lot of current, I don't want any electrical short happening with that many amps available. The wires are insulated and rated to carry the current of that diode, but I want two forms of protection just in case something gets too hot.

I've used Bond Ply 100 and TGF120 for the heat sinking of the diode and driver. As you may or may not know the SST-90 can get really warm and building it as I described above I've never had a diode fail on me, nor get that hot in operation. Keep in mind my lights operate under water so they cool off very well, but I do need heat transfer from the diode to the water, and it works well.

I've used them out of the water but the SST-90 in a handheld light gets hot and after a while you just don't want to hold it. For diving it's perfect.

IME, the real world, anodysing is perfectly fine. My coatings are .5 mils thick. For what it's worth I could mount the LED direct to the button but I prefer the thermal pads for an added layer of protection. They really do work well and you can remove the diodes easily if you needed to down the road.

As to the numbers, the better the conductivity the better it is, but there is only so much mass present in the flashlight and you'll find that while technically it's better in reality the performance is the same. It's a game of tolerances and specifications of what the LED needs to survive. Exceed it's tolerance and it dies, keep it less and it lives. So why spend the time and money for the best you can buy and make if it's of no significance to the real world operation of it?
 
Last edited:
For what it's worth I could mount the LED direct to the button but I prefer the thermal pads for an added layer of protection. They really do work well and you can remove the diodes easily if you needed to down the road.

I completely agree with your assertion- you should do those things that make sense to do. And certainly a dive light is a different beast than a flashlight.

I am curious though- when you say you use a thermal pad, are you saying that you are using the material to affix the LED's star board to the heat sink? If so, how hard are you driving the SST-90?
 
Pretty much. I would suggest you re-examine your diode mounting strategy though.

First off, AS5 conducts at about 10 (maybe) where the worst of solders (50/50) conduct at 20. 97/3 conducts in the low 60's and indium in the 70's. But let's not put the cart before the horse.

AS5 won't give me a bond by itself. I was considering using the thermal adhesive: http://www.arcticsilver.com/arctic_silver_thermal_adhesive.htm

I don't have values for conductivity of the adhesive, but it would have the advantage of being able to mount to the (more conductive) non-anodized aluminum. If it's the same as AS5 (~10), then I suppose that would be the bottleneck.

Are you saying that the worst solders conduct at 20, and this would be the effective bottleneck before the anodized aluminum layer (at 30)?

You need to look at your thermal path as a part of your overall design.

You've been more than willing to put a ton of effort in your responses, so I can at least clue you into my design.
I'm going for a tiny palm sized rectangular light, custom CNC cut from aluminum plates. It will house 2 XML-2 U2 emitters and a single XPG-2 R5. It will have adjustable brightness and can turn on the XML pair and XPG thrower independently. The battery will be a newer lithium polymer cell with a C rating capable of reliably running all 3 emitters at max current if I want, which is a potential of 7.5A. Blasting all three at max power would only net a ~7-10 minute run time at 100% DOD, at a serious >2000 Lumen. That is worst case, but in daily situations would probably do 0.5A to each XML for a nice flood and >1 hr run time. The heat transfer issue is of concern to me because the better I design it, the longer I can run at higher power levels.

I realize a (classic) radial design is best for heat dispersal, but I want this to be a low profile pocket light. The power output, body thickness I cut, and driver efficiency are all going to play a huge role in this. If I must, I'm willing to put in micro-fans and maybe some heat pipes/fins if the body saturates quickly. Should I bother anodizing the exterior? I don't care about waterproofing.

I have the CNC machine, aluminum, small optics, the batteries, diodes, etc. It's slowly coming together.

The smaller the die, the smaller the amount of energy flow. The warmer the diode, the less light it emits and the more input energy is turned into heat. You will eventually reach the point where the diode won't go any further. But getting there, the additional energy is simply going to waste. That's why when you put a 1000 lumen and an 1100 lumen flashlight side by side, you're hard pressed to identify which is which.

Yeah I know that perceived brightness is pretty much logarithmic, and brightness of a LED is somewhat inversely related to the temperature. We're always fighting the law of diminishing returns. I have read anecdotal claims of 3% brightness lost for every 10C temp increase. Basically, I will be considering fan cooling if power consumption lost via fan more than makes up for efficiency lost from warm LED's. I have a strong inclination that the mass limitation alone will necessitate active cooling. I'm probably not going to run the LED's much higher than their defined max ratings, so I suppose I should focus more of my attention on convecting away the heat from the body. Thoughts?

PS: thanks for the info on the copper cored MCPCB's, I always like options!
 
In such a small light, I believe you'll quickly run into diminishing returns in passive thermal design. That is, regardless of the clever thermal junctions (the real enemy to thermal conductivity of the light) you will saturate the light body too easily. The LED itself has about 4C/W thermal resistance (check the data sheet) so at a reasonable 5W it will be 20C warmer than the light. The larger interfaces of the LED thermal pad and mounting tend to have lower resistances, so let's estimate LED temperature as being about 40C warmer than the surface of the light. The LED is rated to 150C, making the light body about 110C at failure. But the battery is probably rated to 60C, and I bet my hand is only good to 55C for a few moments. With good thermal design, reasonably driven LEDs probably aren't the thermal limit.No small light will run for long without active cooling. I find P60 style lights to get uncomfortably warm with over 4W waste heat. Smaller lights like your plan have lower continuous-run limits. There are IP68 fans in the world, but they are not cheap.7.5A at around 3.5 v is 25W. My big P60 lights are too hot to hold soon at 15W. I think you'll reach unsafe temperatures within a minute (unsafe to hold) without active cooling. Plug in a 25W soldering iron and hang on... You'd heat a 100g light 15-20 C per minute at max, with minor heat flow to the hand holding it. From 20C ambient to 50C danger zone is just a few minutes. Thermally speaking, you could reject that much heat. But the LED will still be rather warmer than your fan's heatsink. This reduces temperature maximums and saves the user... But 25W of LED is quite a lot.
 
Last edited:
I mis-spoke on the AS5. I was thinking adhesive and wrote compound.

I think I have a good idea of what you're trying to build. From a thermal perspective, I think you have a challenge in two areas. First is conducting heat away from the diodes and second is removing that heat from the flashlight's body. You can do the math, but I think you're going to find that the amount of energy you end up turning into heat is going to saturate your mass very quickly- long before you run out of battery. So to answer one of your questions, I don't think anodizing the exterior is going to do much for you, simply because of a lack of surface area.

I have a little experience with flashlights, a pretty decent level of experience with heat management, but I've been working on making things that shouldn't work actually work and work well for about 35 years. Experience has taught me three things. The first is to always look for the simplest solution. Second is to do (or not do) everything for a quantified reason- let the facts and data make your decisions for you. Third is that I never fall in love with any of my ideas. That one prevents you from being blinded to your next, better idea.

These are some random thoughts and observations-

You're going to be driving these diodes above their most efficient levels. That means you're turning battery energy into heat output instead of light output. I would strive for a design that was the most efficient. Second, I would look to simplify the design. Is there an approach that would allow you to use one diode? If so, that would likely free up room inside. Room for things like enhanced cooling. You write of three diodes, using one for throw and two for flood. My limited experience tells me that without a large and deep reflector, throw isn't going to be all that great. I might suggest looking at a single, large diode. Maybe a Luminus CBT-140 or a Cree MT-G2. These would give you a TON of flood while driving them at their efficiency sweet spot. maybe consider using some of your new-found internal space for a moveable optic. Maybe have a TIR that could slide laterally in front of the diode for your throw application.

On the heat pipe, all you accomplish there is moving heat from one place to another. You have a very small device that lacks a space away from the heat generation source that you can stash this excess heat. What a heat pipe might do is allow you to reduce your overall mass- or use that weight in a more efficient fashion. As I said, in that size, if you made it of solid copper you'll still saturate quickly. Maybe you mount your diode on a heat pipe and the other end is finned and in the airflow of a small fan.

As far as efficiency losses as the diode warms, you'll find that this isn't going to be a liner curve. The closer you get to the diodes thermal limitations, the more impact each degree makes. As heat rises, efficiency drops, which only serves to cause more of your battery power to be converted to heat instead of light which raises the temperature up some more which reduces efficiency... Sort of a hand-held China Syndrome.

I would focus on efficiency and heat removal. Instead of having a set idea on what you want to build, I would have a grouping of individual considerations. You can massage and tweak each one until you arrive at a final design. Make one your package size and configuration. One would be power supply. One would be light output volume and type. Every one of them is going to have an ideal- and that will be something that's likely unachievable. You'll have to trade compromises between each column until you have an acceptable balance that gives you a nice product- or at least tells you what you want to achieve at the minimum level of compromise isn't a practical product.
 
For reference, my daily light right now is a Shining Beam S-Mini: http://www.shiningbeam.com/servlet/the-208/ShiningBeam-S-dsh-mini-XM-dsh-L-T6/Detail
The whole body only weighs 1.6 Oz and can run 1A continuous drive current while tail standing, without overheating. This gives me a baseline idea of what is capable.

Like I said, most of the time I will be driving this custom build around 0.5A each to the 2 XML's. Running two together increases efficiency at lower drive currents and doubles the contact patch for heat dispersal. When analyzing the XM-L2 datasheet's graph of current vs luminus flux, it's clear that running two at 0.5A will easily put out more light than a single XML at 1A, all other things being equal. Since the XML's are used just for flood, surface area is irrelevant, and it doesn't really matter how many LED's it comes from. When I want throw, I'll likely turn on the single XPG alone at 1.5A. I purchased extremely small aspheric lens optics for this; if those don't work due to inadequate focal range, I think I might hunt down a fresnel lens. Do they even make single ultra small TIR's for XPG? I've always had a thing for the "zoomable" lights, but I figured this feature would be irrelevant if I could swap between turning on flood vs pre-focused spot using different LED's in the same light.

My interest in the best thermal design is purely to maximize the "wow" output time on occasions when I want to absolutely blast everything at max output, which won't be often. The better I make it, the more continuous current I can use before hitting an unacceptable temp range. I'm sure you can agree the 2x XML is better for my purposes now that I've explained this. Correct me if I'm wrong; you seem to know your stuff! For thermal management, the smallest copper heatpipe I have been able to find is 100mm (4"). I'd like to be able to get it down to 3", but no luck there. From there, I will work on fan design. I have micro fans as small as 0.6"^2 I'm willing to experiment with. I've got a decent amount of design and testing to do. I didn't even mention yet that this is going to be USB rechargeable :thumbsup:

I'm having a hell of a time finding an adequate driver. I'm really trying to keep it a single cell (3.7v nominal) buck/boost design to keep a simple charging circuit. I don't want to design my own driver if I don't have to. In all likelihood, I'll probably have 1 driver for the 2 XML's in series and 1 driver for the XPG. Easily soldered connections for pots would be a plus. I'm willing to sacrifice max output, especially if the size limitations don't make heat dispersal remotely feasible. Do you have any suggestions? Since I'm likely never going to buy another light after this, price doesn't matter much here.

You guys have been awesome so far for ideas, and I really appreaciate it.
 
I'm by far no expert so whatever I tell you should be taken in this context. I'm learning just as many others are. In this instance some of what you're doing is coincidental to something I'm working on.

The reason I suggested that you look at the MT-G2 is that I think you'll find that it's going to be a bit more efficient than your 2x XM-l2 setup while providing for a cleaner execution. If your goal is to have a regular-use mode at about 300lm and a wow mode, the MT-G2 is certainly capable of the latter. It also has a much larger contact patch, which may be a detriment- it's a bit larger than the XM-L. But 13sq/mm vs 53sq/mm is 2x the thermal path of a pair of XM-l's. On the MT-G2, my thought would be to have your aspheric mounted so that it could be moved in and out over that diode. Again, I'm thinking about how to get more into the package or make the package smaller. One diode eliminates additional switching, wiring and sinking. The fresnel lens is a non-starter. They require a substantial focal length that you're not going to have room for. You might want to inquire about the Wavien focus ring donut thingy. I have no experience with it and it may just be a glorified TIR, but it could be interesting. The technology has gone un-noticed for some time now, but just recently a respected builder announced an affiliation with the company. So maybe there's something to it.

All the heat pipe is going to gain you is moving heat to a more advantageous place to then remove it. I'm not sure you'd gain much in this design. With this additional information and some further thought, I'm imagining that your high output level thermal management strategy is going to end up being delaying the onset of thermal saturation for as long as possible when in wow mode. What you're going to run into is that the light will still be operating within parameters but too hot to hold. You might reconsider using aluminum. If you were to construct the body of copper, you'd have quite a bit more "buffer" for saturation. In terms of providing insulation while maintaining aesthetics, you might consider skinning the unit in titanium. Obviously this is going to require active cooling, but I doubt there's any way to escape this. From a thermal emission perspective, the barrier is going to be surface area. One area where aluminum excels is in emission. If heavily anodized, it will far and away out-emit copper. You may run into a challenge here in getting enough surface area inside this small light. You may have to construct (or find) a foil-thickness heat sink.

I love the "last light" comment. If that holds true, you'd be a rarity. Me- I get to thinking about the next one about the same time I transition from planning to building the present one. Seems like I'm always stumbling across a material, item or approach that may not work for what I'm doing now, but is something I'd like to play with.
 
The thickness of the anodization doesn't have much of an effect on thermal conductance. Ironically, thickness does have an effect on emittance. Anodizing has a dramatic positive effect on aluminum's emissivity.

Your conductivity numbers aren't wrong, but they don't really translate exactly to the specific situation. The ability of alumina to conduct heat is only one part of a 3-dimensional puzzle. The first thing that must be considered is that heat conducts through matter in two fashions- in-plane and through-plane. It's pretty much as it sounds- in-plane is how it moves through a single mass and through-plane is how it moves between two pieces of material. As I have written about before, you could take two pieces of the identical material and mate them together as closely as possible and they won't conduct the same as the single solid mass.

Heat is the energy of excited electrons. Fourier's law states that heat moves towards cold.

Heat must be more than the energy of excited electrons. In a metal for instance, the only electrons that can get excited from absorbing kT thermal energy are the ones near the Fermi level. The others can't absorb any energy because moving up in energy state would violate the Pauli exclusion principle, i.e., the electron states where these excited electrons are trying to go are already occupied. The states above the Fermi level are unoccupied, so electrons near Ef can move up from thermal excitation. So most of the electrons can't participate at all.

There is also heat from phonons, and for gases, heat from atomic and molecular collisions (i.e., kinetic theory of gases).

Fourier's law more specifically says that heat flows in the direction of the negative thermal gradient. Thus, the direction of heat flow is perpendicular to the temperature isotherms. For the boundary condition of a constant heat flux at the surface, the thinness of the anodized layer matters because it affects the temperature profile that you get. For a thin layer, the temperature profile in steady state is essentially flat and it reaches that steady state faster than if you were looking at heat conduction in a thick slab. It's analogous to why you want a very thin layer of thermal paste or thermal adhesive under an LED -- thermal resistance is proportional to the bond line thickness.
 
I have seen this done in order to get the bare emitter on solid aluminum. Since it's so difficult to get the solder to attach, I have seen people anodize the area and it accepts reflow solder better. However, anodizing.... or adding a layer of aluminum oxide to the surface, lowers the thermal conductivity from 237 W·m−1​·K−1 ​to 30 W·m−1​·K−1​. This is bad... like nearly as bad is Titanium bad (which is rated at 21.9).

If the whole point of direct mounting a bare emitter to a heatsink is to bypass the inefficiency of the star, why would anyone do this? Are my numbers wrong?

I'm not clear why anodizing improves solderability. The anodized layer is still aluminum oxide, which is the same as the native oxide layer. The reason I've seen anodizing done when mounting certain LEDs is that those LEDs (e.g., the old Seoul P4 and P7) use a slug that is connected to the anode. Thus, you need to isolate the slug electrically from the heat sink and anodizing is a simple way to do that, and also affords some level of redundancy if your electrically insulating thermal glue (e.g., Arctic Alumina epoxy) is too thin in spots.
 
Heat must be more than the energy of excited electrons. .

What I hope to do is to provide some practical insight into how heat works. For that to be helpful, I've tried to reduce it to the most basic of concepts.

My thinking is that if the average builder picks up a few tips or learns more about how heat moves, they can put that information to good use. For example, I have seen more than a couple times the use of brass in heat sinks. Logic would cause one to think that brass, being mostly copper, would conduct heat about the same when in reality brass is a very poor thermal conductor.

I hope my comments don't come off as "look how smart I am..." as that's not how they are intended. As I've posted before, I'm a high-school grad with just a bunch of hands-on experience. Nothing more. I'm not even sure I understand 100% of your post. I do know that if the conversation gets too complicated or too far over the average guy's head, they stop reading. I know that as I'm one of those guys. When that happens, information that might be useful is never communicated.

If you have a basic understanding of thermodynamics, what I've been posting isn't intended for you. But as I've written before, seeing some of the things I have leads me to believe a few tips might be helpful. When a hobbyist builds a light, they plan in great detail the needs of the physical design, they do the numbers on power supply needs and in some cases do optics calculation. My hope is that this discussion shows the builder that heat management can be quantified and calculated, if only in comparative terms.
 
I do know that if the conversation gets too complicated or too far over the average guy's head, they stop reading. I know that as I'm one of those guys. When that happens, information that might be useful is never communicated.

I'm definitely not one of those guys. I love it when people use terms I don't know. I either ask or look them up when I encounter them; not everyone has taken certain engineering and physics classes. How else do you learn new things??

BTW I've been looking over the MT-G2, and it does appear more efficient over the entire range even when comparing to 2x XM-L2. Unfortunately, I already purchased the XM-L2's and XPG2's. If the design necessitates it, i'm willing to invest in this diode, but it doesn't appear to have a top that I can solder. It would necessitate a star board. I'm willing to go that route if I have to; options are always nice. I'll be slowly working on this, as I am still a few essential parts short anyway.
 
One of the reasons i mentioned the MT is that it has a much larger thermal pad. This could be attached directly to a copper sink with the + and - tabs free floating while still giving you more thermal contact area than the pair of XML's. If you're going to move forward with the XML, I would suggest looking into using a cut-down to fit copper star with a direct-contact center pad. You still may run into mechanical issues in the implementation. Doing the numbers, if you have a diode base contact area of X that's soldered with a connection at 60 W/m K and it's conducting through the electrical insulating barrier at 2 W/m K, your star needs to be 30 times the footprint. If you go direct to a copper sink with a 60 solder joint, you'll still need at least six times (I'd estimate 10+) for the epoxy bond to your sink.
 
Top